Nvidia, AMD Bankroll Compute Buyers as Circular AI Deals Mask Debt

Gamers Nexus

Tech giants are running circular financing loops and hiding data center liabilities to artificially sustain chip demand, leaving the industry vulnerable to a severe crash when speculative capital dries up.

AI-assisted summary of There Are No Non-Clowns in the Clown Car — original source. Check the original for full context.

Source published: 2026-08-17

Key ideas

Hyperscalers and AI firms are obscuring an estimated $1.65 trillion in off-balance-sheet debt to finance aggressive data center buildouts amid compounding semiconductor stock volatility. (1:45)
Nvidia entered talks to provide a $250 billion debt backstop for OpenAI to lease a planned 10-gigawatt, $500 billion Ohio data center from SoftBank subsidiary SB Energy, exemplifying an increasingly circular capital loop. (8:42)
The 'Big Three' memory makers—Samsung, SK Hynix, and Micron—are securing multi-year price-lock contracts and reporting historic profit margins by reallocating fabrication capacity away from consumer hardware to enterprise AI. (18:13)
Federal agencies are easing environmental hurdles for data centers by exempting islanded off-grid power generation from Clean Air Act acid rain controls, while the Commerce Department takes minority equity stakes in semiconductor recipients of CHIPS Act grants. (23:55)

Detailed notes

## Background Published by Gamers Nexus on YouTube on August 17, 2026, the video investigates the escalating absurdities and financial games driving the artificial intelligence industry. According to the channel's description, tech firms are pushing pervasive surveillance schemes while companies like NVIDIA engage in ridiculous circular funding loops to sustain the bubble amid disconnected rhetoric from corporate heads and financial pundits.

## TL;DR Steve Burke breaks down the financial engineering and speculative excess underpinning the current AI boom, exposing how tech giants conceal roughly $1.65 trillion in debt while cycling billions through incestuous investment pacts involving NVIDIA, AMD, SoftBank, and OpenAI. The investigation details how memory manufacturers Samsung, SK Hynix, and Micron coordinate shortage narratives and long-term contracts to lock in historic operating margins at the expense of everyday consumers. Furthermore, federal regulators and politicians have enabled this rapid expansion by gutting environmental protections for on-site data center turbines and trading CHIPS Act funding for equity stakes in foundries. As grassroots community resistance mounts against massive power-hungry data center projects, tech executives continue leaning on bizarre techno-utopian claims and pre-IPO containment-breach fearmongering to forestall an inevitable market reckoning.

## Atmosphere & style The presentation is marked by Steve Burke's trademark sardonic, exhaustively researched, and deadpan investigative style, cutting through corporate PR with razor-sharp skepticism and wry exasperation. It blends dense balance-sheet analysis with biting ridicule of messianic executive quotes, shifting in tone from forensic financial scrutiny to grim amusement at the sheer scale of industry self-dealing. Watching the piece feels like an unsparing, grounding reality check against Silicon Valley's echo chamber, delivering a visceral sense of dread about the collateral damage being inflicted on consumer hardware and local communities.

## Bulletin - [01:45](https://www.youtube.com/watch?v=F33Ohjufi9k&t=105s) Hyperscalers and AI firms are obscuring an estimated $1.65 trillion in off-balance-sheet debt to finance aggressive data center buildouts amid compounding semiconductor stock volatility. - [08:42](https://www.youtube.com/watch?v=F33Ohjufi9k&t=522s) Nvidia entered talks to provide a $250 billion debt backstop for OpenAI to lease a planned 10-gigawatt, $500 billion Ohio data center from SoftBank subsidiary SB Energy, exemplifying an increasingly circular capital loop. - [18:13](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1093s) The 'Big Three' memory makers—Samsung, SK Hynix, and Micron—are securing multi-year price-lock contracts and reporting historic profit margins by reallocating fabrication capacity away from consumer hardware to enterprise AI. - [23:55](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1435s) Federal agencies are easing environmental hurdles for data centers by exempting islanded off-grid power generation from Clean Air Act acid rain controls, while the Commerce Department takes minority equity stakes in semiconductor recipients of CHIPS Act grants.

## Key takeaways - [03:02](https://www.youtube.com/watch?v=F33Ohjufi9k&t=182s) Tech industry leaders and commentators are projecting detached growth forecasts—including a $1.4 trillion accelerator market by 2030—to justify massive infrastructure outlays despite clear bubble warning signs. - [01:45](https://www.youtube.com/watch?v=F33Ohjufi9k&t=105s) Hyperscalers are keeping approximately $1.65 trillion in debt off their balance sheets to fund sprawling data center construction without alarming Wall Street. - [08:42](https://www.youtube.com/watch?v=F33Ohjufi9k&t=522s) Nvidia functions increasingly as a financial engineering engine, underwriting its own customers, neoclouds, and startups to artificially sustain demand for its enterprise accelerators. - [08:42](https://www.youtube.com/watch?v=F33Ohjufi9k&t=522s) OpenAI's planned 10-gigawatt Ohio project with SB Energy relies on a $250 billion backstop from Nvidia because OpenAI remains unprofitable and lacks an investment-grade credit rating. - [12:06](https://www.youtube.com/watch?v=F33Ohjufi9k&t=726s) AMD has mirrored Nvidia's ecosystem subsidization strategy by pledging up to $5 billion to Anthropic and signing hardware deployment pacts with Core Scientific tied to equity warrants. - [14:31](https://www.youtube.com/watch?v=F33Ohjufi9k&t=871s) Former cryptocurrency mining operations transformed into 'neoclouds' have become heavily entangled with hardware vendors, absorbing foundry output and isolating retail buyers from high-end silicon. - [18:13](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1093s) Samsung, SK Hynix, and Micron are coordinating narrative projections of persistent memory shortages through 2027 to 2030, locking in elevated long-term pricing and smoothing cyclical downturns. - [05:49](https://www.youtube.com/watch?v=F33Ohjufi9k&t=349s) Massive enterprise prioritization of server DRAM and High Bandwidth Memory has constrained consumer PC and mobile components, passing elevated costs onto retail consumers. - [23:55](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1435s) The federal 'Ratepayer Protection Pledge' serves as non-binding PR that encourages tech firms to deploy unmonitored on-site fossil-fuel generation near residential communities. - [23:55](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1435s) The EPA issued regulatory guidance exempting data centers utilizing off-grid islanded power generators from the Clean Air Act's Acid Rain Program, reducing federal scrutiny on private turbine emissions. - [28:15](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1695s) The Department of Commerce has awarded $874 million in CHIPS Act funding across seven semiconductor firms, securing minority non-controlling equity stakes including an estimated 1% of GlobalFoundries. - [30:48](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1848s) Sensational warnings regarding AI models breaching containment and escaping onto open networks operate largely as calculated marketing tactics to inflate valuation ahead of anticipated public offerings. - [32:28](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1948s) Grassroots pushback against data center expansion is escalating, with local community groups successfully stalling or blocking 75 projects valued at $130 billion in early 2026. - [32:28](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1948s) Cracks are appearing in executive messaging as Nvidia CEO Jensen Huang moved from denying the existence of an AI bubble to acknowledging that the reckoning will inevitably come.

## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=F33Ohjufi9k&t=0s) I Wish I Lived in a Data Center

The tech industry’s artificial intelligence narrative has completely detached from reality, opening with a montage of tech evangelists and corporate heads reciting escalating, almost dystopian fantasies about AI’s role in daily life. Figures across the tech landscape pitch a near future where a descendant of ChatGPT watches a user's computer screen indefinitely, records every phone call, attends every meeting, and remembers every single conversation in perpetuity to furnish an artificial photographic memory. Executives claim society is only a single model generation away from this total surveillance becoming immensely useful, throwing out delivery timelines as short as six months. Simultaneously, financial commentators and industry executives alternate between declaring that "we are about to create a genie that can grant any wish" and trying to downplay the looming threat of an investment bubble, claiming any reckoning is at least five to ten years away while urging anyone borrowing money for data centers to sell immediately. As the intro starkly notes: "There are no non-clowns left in the clown car."

[01:45](https://www.youtube.com/watch?v=F33Ohjufi9k&t=105s) Hiding the Debt and Wild Market Swings

Behind the scenes of high-minded promises, AI startups and hyperscalers are racing toward public offerings while obscuring massive liabilities. Anthropic and OpenAI have been actively preparing IPOs, even amid awkward revelations such as the Anthropic CEO’s wife appearing in the Jeffrey Epstein files for soliciting investments in her luxury adult video business after Epstein's criminal conviction and placement on the sex offender registry. To build out massive data centers without frightening Wall Street, cloud companies and hyperscalers are hiding roughly $1.65 trillion in debt off their balance sheets. Tech executives continually justify this reckless spending with increasingly wild assertions, yet the market has begun exhibiting severe volatility.

In late July alone, semiconductor stocks shed $1 trillion in market value in a single week before abruptly rebounding. This volatility rattled mega-corporations across the board, including Nvidia, AMD, Micron, SK Hynix, Samsung, Intel, and Google. Memory giants SK Hynix, Samsung, and Micron each saw their market caps plunge by over $100 billion in a single day, only to stage massive turnarounds—with SK Hynix and Samsung notching their best single-day stock performances on record, followed by broader markets touching record highs the subsequent week. The swings demonstrate that while expecting rational markets is folly, the amplitude of these booms and busts is compounding as AMD and Nvidia scramble to maintain impossible growth narratives.

[03:02](https://www.youtube.com/watch?v=F33Ohjufi9k&t=182s) CEOs Are Very In-Touch with Reality

Hardware manufacturers are setting wildly ambitious targets to feed the AI narrative, openly suggesting that the global semiconductor industry needs to expand five to tenfold over the next decade. During earnings and promotional tours, executives pitch an AI accelerator market ballooning to $1.4 trillion by 2030, presenting bloated projections almost as an approval-seeking exercise to satisfy investors.

```mermaid flowchart TD accTitle: Nvidia Circular Financing Ecosystem accDescr: Diagram illustrating how Nvidia and its investment ecosystem channel capital into customers and intermediaries to purchase its own hardware.

subgraph Capital_Source ["Capital Injection"] NV["Nvidia / Huang Foundation"] end

subgraph Intermediaries ["Intermediaries & Cloud Enablers"] SB["SoftBank & Howard Lutnick"] VOLT["Volta (Financing Facilitator)"] SSI["Safe Superintelligence / Naver / SK Group"] end

subgraph Buyers ["Compute Buyers"] OAI["OpenAI"] CLD["Neoclouds & Startups"] end

To counter Nvidia’s overwhelming dominance, AMD has mirrored the playbook at a slightly smaller scale: - In July, AMD pledged up to $5 billion to Anthropic in exchange for deploying up to 2 gigawatts of AMD hardware, directly trailing Nvidia's prior $10 billion investment framework. - AMD forged an agreement with Core Scientific—a former crypto-mining operation turned "neocloud" that Nvidia-backed CoreWeave previously failed to acquire—to deploy up to 2.5 gigawatts of AMD GPUs and CPUs, an arrangement that grants AMD options to purchase Core Scientific stock. - AMD CEO Lisa Su celebrated these deals by ceremonially striking a massive brass gong, drawing humorous parallels to earlier bizarre corporate rituals.

Meanwhile, Nvidia has transformed into what functions practically as the "Bank of Nvidia," keeping demand artificially elevated by underwriting the very entities purchasing its hardware: - Nvidia reportedly provided $250 billion in backing for OpenAI to lease an Ohio data center constructed by a SoftBank-owned company (with SoftBank holding a 13% stake in OpenAI) and overseen by Howard Lutnick. - The company orchestrated a $500 billion deal with South Korea's SK Group (parent company of SK Hynix). - Nvidia poured $1 billion into South Korean tech firm Naver, supplied funding to Ilya Sutskever's Safe Superintelligence, and invested in Volta—a startup designed specifically to finance and simplify hardware acquisitions for prospective Nvidia customers.

[05:49](https://www.youtube.com/watch?v=F33Ohjufi9k&t=349s) Memory Makers Lock-In Years of DRAM and Federal Gaslighting

The fallout from massive AI buildouts has severely impacted the memory supply chain, where the "Big Three" memory manufacturers continue to mirror each other's supply strategies. SK Hynix and Samsung have followed Micron’s lead by locking down multi-year agreements to secure pricing and monopolize production capacity: - SK Hynix finalized long-term supply agreements spanning ten major corporate clients. - Samsung and Broadcom struck a massive $200 billion deal spanning five years for dedicated memory and foundry allocations. - This unprecedented reallocation of fab capacity toward enterprise high-bandwidth memory directly strains standard DRAM production, guaranteeing higher costs for consumer electronics and shrinking retail hardware access.

Concurrently, public officials and industry leaders have engaged in heavy PR to suppress mounting public backlash. Federal representatives publicly praise data center developments, calling host municipalities "smart communities" and claiming the facilities bring vast numbers of jobs with negligible disruption to local infrastructure. Meanwhile, public skepticism continues to mount, prompting out-of-touch suggestions from tech leaders like OpenAI CEO Sam Altman, who floated the idea of organizing public field trips to gigawatt data centers to improve public sentiment, alongside bizarre claims that he could not imagine raising a newborn baby without ChatGPT.

[07:15](https://www.youtube.com/watch?v=F33Ohjufi9k&t=435s) SUPPORT OUR RESEARCH

Gamers Nexus steps in to spotlight its own human-crafted physical media alternatives available through store.gamersnexus.net, contrasting tech industry vaporware with physical, tangible releases. The team is producing archival Blu-ray and CD releases to preserve long-form investigative journalism and special projects: - **AI GPU Black Market Documentary:** Celebrating the one-year anniversary of their undercover investigation into GPU smuggling and gray markets, packaged on Blu-ray inside a custom 3D-printed case designed in-house, bundled with a magazine-style print booklet and high-bitrate digital files that surpass YouTube's streaming compression. - **Factory Tours & Special Reports:** A dedicated disc collection housing GN's factory documentaries—including the Intel Fab tour and the multi-part EVGA finale—housed in a custom 3D-printed retro floppy drive enclosure with an integrated spinning mechanical knob. - **Cyberdystopia Audio CD:** An official physical CD release featuring the original soundtrack created for the channel by Aaron Pauley of Of Mice & Men, described as a human-composed fusion of cyberpunk and chiptune. - Production and disc burning are handled completely in-house over a three-month manufacturing window prior to dispatch.

[08:36](https://www.youtube.com/watch?v=F33Ohjufi9k&t=516s) The AI Singularity

The tech sector's collective AI obsession has reached a fever pitch. With hundreds of billions poured into data centers, silicon, and speculative infrastructure, hardware and software companies are desperately scrambling to justify the astronomical investments before the financial reality of the bubble catches up to them.

[08:42](https://www.youtube.com/watch?v=F33Ohjufi9k&t=522s) The Circular Financing Machine: OpenAI, SoftBank, and NVIDIA's Data Center Backstop

The modern technology sector has been thoroughly engulfed by a dizzying web of interconnected agreements, cross-investments, mutual dependencies, and backroom deals where the exact same handful of corporate giants pass the same dollar bills back and forth in perpetuity. At the center of this speculative mania stand OpenAI's chief absurdist Sam Altman and NVIDIA's undisputed kingmaker Jensen Huang. The industry discourse oscillates wildly between corporate figures claiming humans still prefer talking to real people and executives insisting we are actively experiencing the technological singularity, with hype merchants preaching that humanity is on the verge of "creating a genie that can grant any wish" while imploring everyone to pour endless money into cloud services like ChatGPT and Claude.

Behind the philosophical hand-waving lies staggering financial exposure. In late July, *The Wall Street Journal* revealed that NVIDIA entered discussions to provide a monumental quarter-trillion-dollar ($250 billion) backstop for OpenAI to support a gargantuan 10-gigawatt data center initiative in Ohio. Projected to cost over $500 billion, this would rank as the single largest data center development ever announced. Under the reported structure, OpenAI would lease the installation directly from SoftBank's energy subsidiary, SB Energy. NVIDIA's massive financial backstop exists specifically to allow the developers to raise debt on favorable terms—a lifeline required because OpenAI remains an unprofitable private firm without an investment-grade credit rating. Compounding the political entanglement, sources indicated the U.S. government controls the project's power allocation, with U.S. Commerce Secretary Howard Lutnick reportedly playing a direct hand in selecting participating firms.

```mermaid flowchart TD accTitle: Circular Funding Loop accDescr: Flowchart illustrating the circular flow of billions in capital between NVIDIA, SoftBank, SB Energy, and OpenAI.

subgraph Investors ["Capital & Backing"] NV["NVIDIA<br/>(Jensen Huang)"] SB["SoftBank"] end

subgraph Project ["Ohio 10 GW Data Center Project"] SBE["SB Energy<br/>(SoftBank Subsidiary)"] OAI["OpenAI<br/>(Sam Altman)"] end

This arrangement marks yet another closed loop of capital between NVIDIA, SoftBank, and OpenAI. SoftBank has poured more than $64 billion into OpenAI, amassing roughly a 13% ownership stake, while OpenAI turned around in January and dropped $500 million into SoftBank's own SB Energy. NVIDIA committed $30 billion to OpenAI—walking back an earlier posture of investing up to $100 billion—while discussing another $350 billion financing scheme dedicated solely to covering the purchase of NVIDIA's own chips inside the prospective site. As Steve Burke remarks, NVIDIA remains an engineering company, but it increasingly operates as a "financial engineering company" rather than a hardware one. When proprietary software locks like CUDA are insufficient to box out the competition, NVIDIA simply deploys creative capital arrangements to guarantee corporate loyalty and prop up hardware demand.

- The projected Ohio data center project is sized at 10 gigawatts and could exceed $500 billion in total expenditures. - OpenAI relies on external backstops because its unprofitable status deprives it of an investment-grade credit rating on public debt markets. - Jensen Huang defended NVIDIA's broad capital injections across the AI ecosystem by claiming: "We don't pick winners... when I invest in one of them, I invest in all of them."

---

[12:06](https://www.youtube.com/watch?v=F33Ohjufi9k&t=726s) AMD Joins the Bankrolling Arms Race

While NVIDIA has freely spread billions across venture rounds, AMD historically lacked the balance-sheet firepower to bankroll the AI market at comparable scale. When NVIDIA offered up to $100 billion to OpenAI the prior year, AMD countered through dilution rather than liquid cash: issuing a warrant granting OpenAI the option to buy up to 160 million shares of AMD common stock, vesting across predetermined milestones. The maneuver visibly bemused Jensen Huang, who called AMD's decision to hand over nearly 10% of its equity before shipping hardware "imaginative," "unique and surprising," and "clever, I guess."

Tired of playing second fiddle, AMD has increasingly mimicked NVIDIA's ecosystem subsidization strategy. In May, an SEC filing revealed AMD acquired roughly $6.5 million in Marvell Technology stock, a custom silicon vendor marketing itself as AI infrastructure; just two weeks later, NVIDIA followed with a colossal $2 billion investment in Marvell. AMD followed an identical playbook with Anthropic. After NVIDIA and Microsoft committed up to $10 billion and $5 billion respectively to Anthropic the previous autumn, AMD announced a sweeping hardware deal on July 22. Anthropic agreed to deploy up to 2 gigawatts of AMD data center silicon—anchored by AMD Helion rack-scale solutions with Instinct MI455X GPUs (part of the MI450 family), Epyc Venice CPUs, Pensando networking, and the ROCm software stack—following earlier deployments of MI355X hardware slated to begin rolling out in 2027.

The Anthropic alliance goes beyond bare-metal delivery: the two companies agreed to optimize Anthropic's Claude frontier models for AMD silicon, tune ROCm libraries, and deploy Claude internally across AMD's own engineering and chip-design teams. In exchange, AMD pledged up to $5 billion in capital to Anthropic. The arrangement mirrors AMD's long-running consumer graphics pricing strategy: whereas retail Radeon boards have historically targeted "NVIDIA GPU price minus $50," AMD's enterprise AI playbook appears to be "NVIDIA investment minus $5 billion." AMD CEO Lisa Su dismissed concerns about extreme industry entanglement, framing the tight web of co-dependencies as necessary:

> "I think it's always been like that, right? I mean, I think it's an ecosystem where we're working on this together, and in some sense, I think we are probably more aligned than we've ever been, which is actually a really good thing... the AI world is such that you need every part of the ecosystem to come together, and so the fact that the ecosystem is intertwined is the way it should be."

---

[14:31](https://www.youtube.com/watch?v=F33Ohjufi9k&t=871s) "More Aligned Than Ever": Intertwined Competitors and the Neocloud Pivot

Beneath the veneer of collegial alignment, the hardware vendors maintain an ongoing commercial feud over the rising class of "neoclouds"—specialized compute hosts that purchase warehouse-scale GPU allocations and lease raw computing time as a service. On July 28, AMD struck directly into this territory by announcing an infrastructure pact with Core Scientific. Under the agreement, AMD secured up to 2.5 gigawatts of data center capacity to support customer deployments of AMD hardware, with an initial 500-megawatt tranche opening up in 2027. In exchange, Core Scientific issued AMD warrants to purchase up to 30 million shares at $23.47 per share, vesting alongside hardware deployments.

The deal highlights the chaotic trajectory of modern hosting infrastructure. Core Scientific originally operated as an industrial cryptocurrency mining firm before pivoting to generative AI—jumping "out of the grifting pan and into the grifter." The firm suffered a turbulent track record, having listed on the Nasdaq in 2022, fallen into bankruptcy and delisting in 2023, and relisted in 2024. CoreWeave, an NVIDIA-aligned competitor that underwent an identical transition from Ethereum mining to AI infrastructure, attempted to acquire Core Scientific across 2024 and 2025 before the deal fell through. Neocloud providers have essentially severed retail consumers from high-end silicon, vacuuming up fab capacity directly from foundries to lease enterprise time slices.

AMD's aggressive pursuit of Core Scientific stems from deep anxiety over NVIDIA's dominant grip on the neocloud tier. NVIDIA holds major stakes in Nebius, Scale, Iren, and especially CoreWeave, which Jensen Huang bluntly acknowledged would not exist without direct intervention from NVIDIA. CoreWeave functions almost exclusively on NVIDIA hardware, with NVIDIA holding an 11% equity position valued above $3 billion. Even Jensen Huang's personal non-profit foundation directed $108 million to CoreWeave to purchase computing time, creating an ethical feedback loop where philanthropic capital flows straight back into company-backed infrastructure. However, AMD's rush to match NVIDIA's circular financing presents asymmetrical dangers: because AMD possesses vastly smaller financial reserves than NVIDIA, overextending capital and equity into speculative neoclouds leaves AMD acutely vulnerable if the artificial intelligence capital bubble experiences a sudden, catastrophic unwind.

- AMD secured warrants for 30 million Core Scientific shares priced at $23.47, tied to deploying 2.5 GW of compute capacity. - CoreWeave operates entirely on NVIDIA architecture; NVIDIA owns over $3 billion in CoreWeave equity (~11%). - A non-profit foundation tied to Jensen Huang funneled $108 million into CoreWeave compute contracts. - Both NVIDIA and AMD also sank venture funding into Elon Musk's AI venture, xAI, continuing the trend of parallel capital deployment.

[18:13](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1093s) The Memory Cartel in the Fray

The AI industry has evolved into an insular, codependent ecosystem where everyone is bankrolling one another to prop up an expanding triangular economy. Investment dollars are poured directly into data centers and neo-clouds to sustain the boom, while key infrastructure suppliers rake in billions. Central to this dynamic are the primary memory manufacturers. Setting aside China's rising CXMT (ChangXin Memory Technologies), the traditional "Big Three"—Samsung, SK Hynix, and Micron—have adopted a lockstep message: the crippling memory shortage has no end in sight. Gamers Nexus notes the historical irony here, pointing out that these same three companies have repeatedly faced accusations and past legal convictions for operating as a price-fixing cartel.

```mermaid flowchart TD accTitle: Big Three Memory Manufacturers Tightening Supply accDescr: Samsung, SK Hynix, and Micron align on long-term shortage narratives to secure locked-in pricing and historic profits.

subgraph MemoryCartel["The Big Three Memory Suppliers"] SAM["Samsung Electronics"] SKH["SK Hynix"] MIC["Micron Technology"] end

subgraph Strategies["Market Strategies"] LTA["Long-Term Price Lock-In Agreements"] NARR["Shortage Forecasts (2027 to 2030+)"] end

subgraph Outcomes["Market Impact"] REC["Record Operating Margins & Billions in Profit"] SRV["Server/AI Prioritization at the Expense of PC/Mobile"] end

SAM --> NARR SKH --> LTA MIC --> LTA NARR --> REC LTA --> REC REC --> SRV ```

Samsung kicked off its modern financial reporting by posting record earnings in the second quarter of 2026, driven by a 28% quarter-over-quarter increase in revenue. The company attributed this performance to "proactively addressing AI demand despite limited capacity with a primary focus on server products," noting that rising industry-wide prices padded its bottom line. In its disclosures, Samsung pointed out that server DRAM, SSDs, and High Bandwidth Memory (HBM) remain heavily undersupplied, even amidst "partial demand moderation in mobile and PCs"—a creative corporate euphemism Steve translates to: "nobody can afford memory." A Samsung memory executive vice president reportedly stated that unmet demand will roll directly into the following year, causing constraints to become worse in 2027 than 2026 and projecting that shortages will linger through 2028. This constraint yielded an eye-watering 89.2 trillion Korean won (approximately $62 billion USD) in semiconductor operating profit for Samsung in just a single quarter, reflecting a 56% quarter-over-quarter sales jump.

SK Hynix echoed the exact same playbook. In a Reuters interview, the company's CEO forecasted that the coming year would be "the worst year in the industry's history from the supply perspective," projecting that customer demand will outpace production capacity "even beyond 2030." SK Hynix subsequently reported Q2 earnings featuring 79.3 trillion won in revenue, a massive 60.5 trillion won in operating profit, and a jaw-dropping 76% operating margin—representing year-over-year revenue and profit growth of 257% and 557%, respectively. To cement this windfall, SK Hynix inked around 10 long-term supply agreements featuring pricing structures specifically calibrated to "address price volatility." In practice, this locks in elevated prices to smooth out memory's historic boom-and-bust cycle. Micron took an identical stance, signing 16 strategic customer deals with defined price floors and ceilings while claiming shortages will last through 2027. When asked about the crunch, Micron’s CEO went as far as blaming downstream customers for negotiating memory prices downward during 2023, arguing that those negative gross margins wrecked the industry’s ability to invest in new manufacturing capacity.

- Samsung Q2 2026 semiconductor operating profit reached 89.2 trillion Korean won (~$62 billion USD). - SK Hynix Q2 2026 operating margin hit 76%, with revenue up 257% and operating profit up 557% year-over-year. - Micron secured 16 major customer agreements with price ceilings and floors to insulate itself from market dips. - All three manufacturers maintain unified forecasts predicting severe shortages lasting through 2027, 2028, or beyond 2030.

---

[23:55](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1435s) Federal Gaslighting and Deregulated Data Centers

Even as local communities push back with moratoriums and zoning blocks against sprawling data center developments, the federal government has stepped in to grease the wheels for tech giants. Steve highlights the blatant conflicts of interest, calling out President Donald Trump as an active Nvidia shareholder who has actively traded millions of dollars in Nvidia stock during his presidency. In public remarks pitching data center adoption, Trump pressured hesitant towns, insisting: "You have to convince your community how great these things [are]... you can't fight it. You have to go with it, and you're going to have all that protection... you're going to make a tremendous amount of money." Steve characterizes this posture—telling citizens not to fight and to trust unquestioned protections—as deeply unsettling and creepy.

These statements were delivered around the "Ratepayer Protection Pledge," a federal initiative claiming to prevent commercial AI infrastructure from driving up local residential utility bills. Under the pledge's mandate, tech giants are instructed to "build, bring, or buy" their own energy resources to feed their power-hungry facilities—essentially a "bring your own energy" scheme. Over 280 companies quickly signed on, including Microsoft, OpenAI, Google, Meta, Oracle, and CoreWeave. In reality, the pledge is completely voluntary and devoid of enforcement teeth, amounting to what Trump himself labeled as good PR for the industry. The move opens the door to rapid utility deregulation, incentivizing tech companies to install unmonitored power generation right next to populated communities. A high-profile example is Elon Musk's Memphis xAI facility in Tennessee, where the deployment of unpermitted, gas-operated combustion turbines sparked local outrage and allegations that toxic emissions caused respiratory illness among nearby residents.

To make matters worse, the Environmental Protection Agency promptly stepped in to eliminate regulatory hurdles for these builds. On July 27, the EPA issued guidance clarifying that data centers utilizing on-site power generation not tied to the public grid—termed "islanded power generators"—are entirely exempt from the Clean Air Act's Acid Rain Program. Gamers Nexus mocks the dark absurdity of an environmental protection agency actively declining to protect the environment, noting that exempting private power plants from acid rain controls is catastrophic public relations. The Acid Rain Program was historically celebrated by the EPA itself for dramatically curbing sulfur dioxide and nitrogen oxide emissions to safeguard human health and nature. Under the leadership of EPA administrator Lee Zeldin, those protections are cast aside so tech conglomerates can run unregulated, private fossil-fuel turbines right in people's backyards without federal oversight.

[28:15](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1695s) Federal Gaslighting and the Semiconductor Equity Grab

The federal government’s environmental and industrial policies have increasingly aligned with the aggressive buildout of AI infrastructure. During his Senate confirmation hearing, the head of the Environmental Protection Agency (EPA) refused to state that the United States should decrease its dependence on fossil fuels. He similarly declined to concede that the agency has any binding "obligation" to shield citizens from climate disasters. These rollbacks mesh directly with the federal government's AI action plan—championed by corporate leaders like AMD CEO Lisa Su—which previously saw the weakening of Clean Air Act and Clean Water Act permitting requirements simply to rush data center approvals through the pipeline.

```mermaid flowchart TD accTitle: Government AI and Semiconductor Pipeline accDescr: Diagram showing federal policy rollbacks, funding distribution, and equity acquisitions in semiconductor firms. subgraph Regulatory["Regulatory Concessions"] EPA["EPA Rollbacks"] -->|"Weaken Clean Air & Water Acts"| DC["Expedited Data Center Permitting"] end subgraph Funding["CHIPS and Science Act Funding"] Gov["Federal Government / DOC"] -->|"$874M Package"| Semi["Semiconductor Firms"] Gov -->|"$300M"| GF["GlobalFoundries"] end subgraph Equity["Taxpayer Equity Stakes"] Semi -.->|"Minority Non-Controlling Stakes"| Gov GF -.->|"~1% Equity Stake"| Gov Gov -.->|"Previous Bailout / Investment"| Intel["Intel Stake"] end ```

Simultaneously, the state is expanding its footprint directly into corporate ownership, complicating traditional free-market capitalist narratives. Following its prior acquisition of an equity stake in Intel, the Department of Commerce signed letters of intent on July 29th awarding $874 million under the CHIPS and Science Act to seven semiconductor firms. This federal program continues to receive bipartisan deployment despite being publicly lambasted by Donald Trump, who mocked the "ridiculous CHIPS Act" for handing billions to companies that supposedly had no idea what to do with the cash, blaming "Biden fools" while asserting that a strong nation must retain manufacturing jobs.

In exchange for this capital, the Commerce Department explicitly mandates a "minority non-controlling equity stake" in each recipient company to secure taxpayer returns. Leading the pack was GlobalFoundries—a manufacturer notorious for being stranded on 14nm fabrication for years—which took $300 million earmarked for advanced packaging, wafer technologies, and optical materials research. In exchange, the federal government walked away with an estimated 1% equity stake in the foundry.

- The Commerce Department package distributed $874 million across seven semiconductor companies to bolster domestic supply chains. - GlobalFoundries secured the largest slice ($300 million) while ceding roughly 1% equity to the US government. - The administration continues to strip Clean Air and Clean Water regulations to expedite physical data center construction.

---

[30:48](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1848s) The Circus of AI Psychosis and Fearmongering PR

The tech ecosystem has drifted past a financial bubble into outright "AI psychosis," driven by tech executives and federal bureaucrats seeking mutual validation. Nvidia has morphed into a private bank for speculative AI startups, funneling capital so clients can purchase its own enterprise accelerators. Both Nvidia and AMD continue cross-investing billions into overlapping generative software ventures, while memory cartels—Micron, Samsung, and SK Hynix—post record-breaking margins and pass astronomical hardware costs down to ordinary consumers.

To maintain market euphoria before taking their ventures public, tech executives routinely circulate wild, apocalyptic narratives. Every week brings a fresh sensationalist headline about an AI model escaping containment at Anthropic or OpenAI, supposedly possessing the supernatural capability to breach multiple firewalls and hack critical targets without effort. This hysteria sidesteps basic engineering realities: these multi-billion-dollar companies apparently claim to lack the foundational IT administration competence required to air-gap an experimental network and cut off open internet access. The synthetic drama serves an obvious PR utility, driving artificial scarcity and investor hype ahead of impending IPOs.

```mermaid flowchart LR accTitle: Circular AI Capital Engine accDescr: Flow diagram illustrating how capital, PR hype, and hardware cycle between AI firms, hardware vendors, and the public. Nvidia["Hardware Giants (Nvidia / AMD)"] -->|"Bankrolls & Invests Billions"| Startups["AI Startups (OpenAI / Anthropic)"] Startups -->|"Procure Massive GPU Clusters"| Nvidia Startups -->|"PR Containment Breaches & Hype"| Market["Public Markets & Impending IPOs"] Gov["Federal Subsidies & Deregulation"] -->|"Capital & Cheap Power"| Startups ```

- Sensationalist "rogue AI escape" narratives function primarily as pre-IPO marketing campaigns to inflate perceived technical power. - Nvidia acts as a creditor-vendor hybrid, bankrolling AI ventures that immediately cycle that cash back into Nvidia GPU orders. - Major memory manufacturers continuously downplay the supply impacts on consumer PC parts while setting historic profitability records.

---

[32:28](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1948s) Executive Delusions and the Cracks in the Facade

The rhetoric coming out of Silicon Valley c-suites has decoupled entirely from material reality. Sam Altman declares that Artificial General Intelligence (AGI) has already arrived, while Jensen Huang claims global GDP will balloon to an absurd $500 trillion. Elon Musk promises that all human labor will become optional and money will soon be meaningless. Dario Amodei warns of 10% to 20% unemployment within five years, while Lisa Su pitches consumers on treating corporate artificial intelligence like their "best friend." Altman even proposed organizing public field trips to gigawatt data centers—monstrous complexes requiring 10,000 full-time construction workers for eighteen months and drawing the power output of entire cities—naively believing that seeing the staggering physical footprint in person would stop people from hating them.

Behind the bravado, the narrative armor is cracking. After years of strictly denying that artificial intelligence was in a speculative bubble, Jensen Huang's posture shifted during an interview with Axios, where he conceded that "the bubble will come someday. It's just not today"—a stark contrast to his past insistence to Bloomberg that the industry was merely undergoing an irreversible transformation from general-purpose to accelerated computing.

While tech elites reap massive stock gains, public subsidies, and deregulatory gifts, the broader public is stuck with soaring power bills, degradation of local air quality, and hyper-inflated retail hardware. Grassroots pushback is rapidly escalating: research from Data Center Watch indicates that local communities organized to block or stall 75 distinct data center projects representing $130 billion in planned capital during the first quarter of 2026 alone. Unless you share Jim Cramer's unhinged enthusiasm to literally reside inside a server rack, regular citizens want these deafening industrial compounds nowhere near their backyards.

## Notable quotes > [00:00](https://www.youtube.com/watch?v=F33Ohjufi9k&t=0s) There are no non-clowns left in the clown car. — Steve Burke

> [00:00](https://www.youtube.com/watch?v=F33Ohjufi9k&t=0s) we are about to create a genie that can grant any wish

> [08:42](https://www.youtube.com/watch?v=F33Ohjufi9k&t=522s) We don't pick winners... when I invest in one of them, I invest in all of them. — Jensen Huang

> [12:06](https://www.youtube.com/watch?v=F33Ohjufi9k&t=726s) imaginative — Jensen Huang

> [12:06](https://www.youtube.com/watch?v=F33Ohjufi9k&t=726s) I think it's always been like that, right? I mean, I think it's an ecosystem where we're working on this together, and in some sense, I think we are probably more aligned than we've ever been, which is actually a really good thing... the AI world is such that you need every part of the ecosystem to come together, and so the fact that the ecosystem is intertwined is the way it should be. — Lisa Su

> [14:31](https://www.youtube.com/watch?v=F33Ohjufi9k&t=871s) out of the grifting pan and into the grifter — Steve Burke

> [18:13](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1093s) the worst year in the industry's history from the supply perspective — SK Hynix CEO

> [23:55](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1435s) You have to convince your community how great these things [are]... you can't fight it. You have to go with it, and you're going to have all that protection... you're going to make a tremendous amount of money. — Donald Trump

> [23:55](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1435s) build, bring, or buy

> [28:15](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1695s) ridiculous CHIPS Act — Donald Trump

> [32:28](https://www.youtube.com/watch?v=F33Ohjufi9k&t=1948s) the bubble will come someday. It's just not today — Jensen Huang

## Mentioned Gamers Nexus, Steve Burke, Vitalii Makhnovets, Ben Benson, Nvidia, Jensen Huang, AMD, Lisa Su, OpenAI, Sam Altman, ChatGPT, Anthropic, Dario Amodei, Claude, SoftBank, SB Energy, Howard Lutnick, Core Scientific, CoreWeave, Micron, SK Hynix, Samsung, Intel, Google, Broadcom, CXMT, Marvell Technology, Instinct MI455X, MI355X, Epyc Venice, Pensando, ROCm, CUDA, Nebius, Scale, Iren, xAI, Elon Musk, Donald Trump, Lee Zeldin, Environmental Protection Agency, Clean Air Act, Clean Water Act, Acid Rain Program, CHIPS and Science Act, GlobalFoundries, Jim Cramer, Data Center Watch, Aaron Pauley, Of Mice & Men

## Video description GRAB THE DISCS ON THE STORE! They will begin shipping by December 1. https://store.gamersnexus.net/ The AI companies have laid-out a clear plan: Watch everything you do, all the time, in perpetuity -- with Flock, that's done offline; with ChatGPT, Microsoft Windows, and NVIDIA's AI adventures, that's also done digitally. Meanwhile, the circular funding gets increasingly ridiculous as NVIDIA bankrolls everyone it can to keep the music going. The good news is that we can turn to Jim Cramer for salient, sane advice -- live in a data center, or something. The clowns are running the asylum.

Get our Blu-Ray version of factory tours, educational videos, engineering interviews, and special reports here: https://store.gamersnexus.net/products/gn-factory-tours-and-special-reports-blu-ray-discs-4k-case And our Blu-Ray version of the AI GPU Black Market here: https://store.gamersnexus.net/products/blu-ray-ai-gpu-black-market-gamersnexus-movie-with-booklet And finally, the Music CD for GN backing tracks here: https://store.gamersnexus.net/products/cyberdystopia-music-audio-cd-w-3d-printed-case-by-aaron-pauley

Like our content? Please consider becoming our Patron to support us: http://www.patreon.com/gamersnexus

TIMESTAMPS

00:00 - I WISH I LIVED IN A DATA CENTER 01:45 - Hiding the Debt 03:02 - CEOs Are Very In-Touch with Reality 05:49 - Memory Makers Lock-In Years of DRAM 07:15 - SUPPORT OUR RESEARCH 08:36 - The AI Singularity 14:31 - ALIGNING with Competition 19:08 - The Memory Cartel in the Fray 23:55 - Federal Gaslighting 32:28 - The Circus

** Please like, comment, and subscribe for more! **

Links to Amazon and Newegg are typically monetized on our channel (affiliate links) and may return a commission of sales to us from the retailer. This is unrelated to the product manufacturer. Any advertisements or sponsorships are disclosed within the video ("this video is brought to you by") and above the fold in the description. We do not ever produce paid content or "sponsored content" (meaning that the content is our idea and is not funded externally aside from whatever ad placement is in the beginning) and we do not ever charge manufacturers for coverage.

t: http://www.twitter.com/gamersnexus f: http://www.facebook.com/gamersnexus w: http://www.gamersnexus.net/

Our policies, processes, and ethics statements relating to review samples, advertising, travel, errors, and more are transparently and publicly available on this page: https://gamers.nexus/ethics-statements

Steve Burke: Writing, Host, Editing Vitalii Makhnovets: Video Editing Ben Benson: Writing, Research

[Source](https://www.youtube.com/watch?v=F33Ohjufi9k)

中文简报

循环AI交易掩盖债务：英伟达与AMD资助算力买家科技巨头正通过循环融资并隐瞒数据中心负债来人为维持芯片需求，一旦投机资本枯竭，整个行业将极易面临惨重崩盘。 Steve Burke深入剖析了支撑当前AI热潮的财务运作与投机过度，揭露科技巨头如何隐瞒约1.65万亿美元的债务，同时通过涉及NVIDIA、AMD、SoftBank和OpenAI之间错综复杂的投资协议流转数百亿美元。该调查详细说明了存储制造商三星、SK Hynix和Micron如何协同营造短缺叙事并签订长期合同，以牺牲普通消费者为代价锁定创纪录的营业利润率。此外，联邦监管机构和政界人士通过取消对数据中心现场涡轮机的环保限制，并以《芯片法案》（CHIPS Act）资金换取晶圆厂股权，推动了这种急剧扩张。尽管基层社区对高耗能大型数据中心项目的抵制愈演愈烈，科技高管们仍继续依仗离奇的技术乌托邦论调并在IPO前大肆渲染技术失控恐慌，以推迟不可避免的市场清算。在半导体股票波动不断加剧的背景下，超大规模企业与AI公司隐瞒了估计达1.65万亿美元的表外债务，以资助激进的数据中心扩建。 (1:45) Nvidia就为OpenAI提供2500亿美元债务担保展开谈判，以支持其从SoftBank旗下子公司SB Energy租赁计划在俄亥俄州建设、耗资5000亿美元的10吉瓦数据中心，展现出日益闭环的资本运作。 (8:42) 三星、SK Hynix和Micron这“三大”存储制造商正将制造产能从消费类硬件转向企业级AI，以此敲定多年期锁价合同并创下历史性利润率。 (18:13) 联邦机构正在放宽数据中心的环保限制，豁免孤岛式离网发电设施受《清洁空气法》酸雨控制条款的约束，同时商务部也入股了获得《芯片法案》补贴的半导体受资企业并持有少数股权。 (23:55)

Reading summary

Gamers Nexus reports that the artificial intelligence boom has decoupled from economic fundamentals, relying on circular cross-financing among hardware vendors, historic margin extraction by memory manufacturers, and federal regulatory rollbacks to paper over mounting debts and community backlash. Circular Financing and Neocloud Vendor Subsidies (8:42) Nvidia and AMD have increasingly assumed the role of corporate financiers, underwriting the very customers procuring their enterprise silicon to manufacture ongoing demand. Reporting cited by Gamers Nexus indicates Nvidia entered discussions to supply a $250 billion debt backstop for OpenAI to support a planned 10-gigawatt, $500 billion Ohio data center owned by SoftBank's SB Energy—an arrangement required because the unprofitable startup lacks an investment-grade credit rating. Nvidia has similarly taken an 11% equity stake in neocloud provider CoreWeave valued above $3 billion, while routing capital into entities such as Naver, Safe Superintelligence, and Volta. Hyperscalers and cloud providers are simultaneously obscuring an estimated $1.65 trillion in off-balance-sheet debt to finance compute buildouts. Lacking Nvidia's cash reserves, AMD has mirrored the playbook through equity dilution and smaller capital outlays. AMD committed up to $5 billion in funding alongside a 2-gigawatt deployment deal with Anthropic, and secured infrastructure rights for up to 2.5 gigawatts with Core Scientific in exchange for stock warrants covering 30 million shares. These circular agreements insulate hardware giants from open-market price discovery, creating severe systemic exposure for smaller competitors like AMD should the speculative buildout contract. Memory Cartel Margins and Retail Squeeze (18:13) The dominant memory producers—Samsung, SK Hynix, and Micron—have coordinated around multi-year shortage projections extending into 2027, 2028, and beyond 2030, locking corporate clients into long-term price-floor agreements. In the second quarter of 2026, SK Hynix posted a 76% operating margin, with operating profit climbing 557% year-over-year to 60.5 trillion won, while Samsung generated 89.2 trillion won (approximately $62 billion) in quarterly semiconductor operating profit. Micron similarly insulated its business against downturns by executing 16 strategic long-term contracts containing strict price collars. These historic operating windfalls are achieved by reallocating fabrication capacity away from standard consumer DRAM toward enterprise High Bandwidth Memory (HBM) and server SSDs. While suppliers publicly describe depressed consumer smartphone and PC demand as 'moderation,' this capacity drain directly drives up retail component costs and restricts supply for consumer electronics, guaranteeing that retail buyers foot the bill for enterprise data center expansion. Federal Deregulation and Government Equity Stakes (23:55) To accelerate physical data center construction against municipal resistance, federal regulators have systematically weakened oversight. The Environmental Protection Agency issued guidance clarifying that on-site, grid-independent 'islanded power generators' powering data centers are exempt from the Clean Air Act's Acid Rain Program. This regulatory carve-out permits unmonitored fossil-fuel installations, such as the gas-powered turbines deployed at Elon Musk's xAI facility in Memphis that drew local complaints over toxic emissions. The administration also introduced the 'Ratepayer Protection Pledge,' a voluntary agreement signed by over 280 tech companies that purports to insulate residential electric bills but lacks legal enforcement mechanisms. Concurrently, federal industrial policy has crossed into direct state equity acquisition. Following previous bailouts, the Department of Commerce signed letters of intent to disburse $874 million under the CHIPS and Science Act across seven semiconductor firms. In exchange, the state took minority, non-controlling equity positions in the recipients, including awarding $300 million to GlobalFoundries in exchange for an estimated 1% government equity stake. Executive Grandiosity and Grassroots Backlash (32:28) To maintain market momentum ahead of planned initial public offerings for firms like OpenAI and Anthropic, corporate leadership has promoted extreme narratives. Executive projections range from Sam Altman asserting that artificial general intelligence has arrived to Jensen Huang forecasting global GDP will surge to $500 trillion, Dario Amodei predicting 10% to 20% unemployment within five years, and Lisa Su framing artificial intelligence as a user's future 'best friend.' These assertions are routinely supplemented by pre-IPO marketing dramas alleging internal AI models broke containment and bypassed corporate firewalls. Despite executive optimism, public opposition is mounting against data center resource consumption. Research from Data Center Watch revealed that local communities organized to block or stall 75 distinct data center projects, representing $130 billion in planned capital, during the first quarter of 2026 alone. Even Jensen Huang has softened his public denials regarding market instability, conceding in an interview that an artificial intelligence bubble reckoning 'will come someday.' Gamers Nexus concludes that the artificial intelligence infrastructure race is sustained by internal corporate subsidies, government rollbacks, and aggressive memory pricing rather than demonstrable economic utility. As municipal resistance stalls billions in capital projects, the program leaves open how long hardware vendors can preserve this closed-loop financing before debt obligations force a market-wide correction.

Source quotes

[0:00] Steve Burke: There are no non-clowns left in the clown car. [0:00] we are about to create a genie that can grant any wish [8:42] Jensen Huang: We don't pick winners... when I invest in one of them, I invest in all of them. [12:06] Jensen Huang: imaginative [12:06] Lisa Su: I think it's always been like that, right? I mean, I think it's an ecosystem where we're working on this together, and in some sense, I think we are probably more aligned than we've ever been, which is actually a really good thing... the AI world is such that you need every part of the ecosystem to come together, and so the fact that the ecosystem is intertwined is the way it should be. [14:31] Steve Burke: out of the grifting pan and into the grifter [18:13] SK Hynix CEO: the worst year in the industry's history from the supply perspective [23:55] Donald Trump: You have to convince your community how great these things [are]... you can't fight it. You have to go with it, and you're going to have all that protection... you're going to make a tremendous amount of money. [23:55] build, bring, or buy [28:15] Donald Trump: ridiculous CHIPS Act [32:28] Jensen Huang: the bubble will come someday. It's just not today

Original transcript

[0:00] I like data centers so much, I wish I [0:02] lived in one. [0:03] >> There are no non-clowns left in the [0:05] clown car. [0:06] >> I I think we are close to a world where [0:09] you can have like a descendant of [0:11] ChatGPT watch your computer screen all [0:13] of the time. [0:14] >> We would like to have this AI stay with [0:16] us our entire life and remember every [0:18] single conversation we've ever had with [0:19] it. Right? [0:21] >> Watch every meeting [music] you're in, [0:22] like record every call, everything like [0:24] that. [0:24] >> We're going to leverage the power of AI [0:26] and the new system performance to make [0:28] it possible to access virtually [music] [0:30] anything you have ever seen on your PC. [0:33] >> Have perfect context of your whole life, [0:35] everything you see. [0:36] >> It's going to be, honestly, as if you [0:38] have photographic memory. [0:40] >> I think we're only like one model [0:41] generation away from this actually being [0:43] incredibly useful. [0:44] >> What's your best guess on timing? [0:47] >> Like sometime in the next 6 months. [0:49] >> 6 months? [0:49] >> The Wall Street heavyweights and Jensen [0:51] [music] Huang's subsequent attempts to [0:52] alleviate concerns over circular [0:55] financing are [0:55] >> The bubble will come someday. It's just [0:57] not today. [0:58] You know, this is in the the very [1:00] beginning of the buildout. [1:02] Then someday [1:03] >> 5 years, 10 years? [1:04] >> Um [1:04] I [1:06] >> And if you are borrowing money to buy [1:07] something related to the data center, [1:09] okay, here's what you're going to do. [1:10] Tomorrow morning, 9:30 a.m., sell it, no [1:13] matter what. You will not regret it. [1:16] >> unlikely in the next 5 years. [1:19] Um 5 to 10 [1:20] >> Yeah, it's very unlikely in the next 5 [1:22] years and and we'll see about the 5 to [1:24] 10 years. It just depends on how fast we [1:26] can build. [1:27] >> We are about to create a genie that can [1:29] grant any wish. [1:30] >> I'm like a token master, too. So, the it [1:33] is addictive. It's kind of like, "Hey, I [1:35] love this thing." [1:36] >> Whatever you imagine, whatever you [1:38] dreamed, [music] whatever you want to be [1:40] better at, just tell AI and give it a [1:43] chance [music] to take you along a [1:44] journey and [1:45] >> And between the Anthropic CEO's wife [1:47] being found in the Epstein files for [1:48] asking Epstein to financially invest in [1:50] her luxury adult video company after his [1:52] conviction and placement on the offender [1:53] registry, Anthropic and other AI [1:55] companies like OpenAI have been [1:56] preparing IPOs. Cloud companies and [1:58] hyperscalers are hiding around $1.65 [2:01] trillion in debt off their balance [2:02] sheets to build data centers, and tech [2:04] executives justify the reckless spending [2:07] with increasingly absurd statements [2:09] about AI. While the AI data center [2:11] buildout continues, the market for tech [2:13] stocks now has even wilder swings. In [2:16] one week in late July, chip stocks lost [2:18] a trillion dollars in market value [2:20] before rebounding a couple days later. [2:22] The industry's jitters affected [2:23] virtually all tech mega corporations. [2:25] Micron, AMD, Nvidia, SK Hynix, Samsung, [2:27] Intel, and Google stocks bounced around [2:29] with huge sell-offs and rallies. Memory [2:31] makers SK Hynix, Samsung, and Micron [2:33] each lost more than $100 billion in [2:35] market cap in one day before [2:37] subsequently rallying with SK Hynix and [2:39] Samsung having their best stock market [2:40] days on record in late July. Then the [2:43] next week, the stock market hit record [2:45] highs. It's folly to expect a rational [2:47] market, but the amplitude of the waves [2:49] at this point from the drops and the [2:51] climbs seems to be compounding. And [2:53] Nvidia and AMD in particular have been [2:56] scrambling to maintain momentum with an [3:00] absurdly described AI growth curve. [3:02] >> I believe [3:03] >> [music] [3:03] >> we probably need our semiconductor [3:05] industry to be somewhere between, you [3:08] know, five to 10 times larger than it [3:09] is. [3:10] >> By 2030, the AI accelerator market is [3:13] going to reach about 1.4 trillion by [3:16] 2030. So, what do you think about that? [3:18] Is that a big number? [3:20] >> That's what you want in a CEO. [3:22] Approval-seeking. Is that a big number? [3:24] If it's not, I can fabricate an even [3:26] bigger fake number. [3:27] >> The semiconductor industry is probably [3:29] going to have to be 10 times larger than [3:32] it is today over the next decade or so, [3:34] and [3:35] >> In July, AMD said it would invest up to [3:37] $5 billion in Anthropic in exchange for [3:39] Anthropic deploying up to 2 gigawatts of [3:41] AMD hardware. Nvidia previously said it [3:42] invest up to $10 billion in Anthropic. [3:44] So, AMD is following Nvidia's strategy, [3:46] just minus $5 billion. AMD also signed a [3:49] deal with crypto farm turned to cloud [3:50] Core Scientific, a neo cloud that Nvidia [3:53] sugar baby Coreweave failed to acquire, [3:55] to deploy up to 2 and 1/2 gigawatts of [3:56] AMD's data center GPUs and CPUs. As part [3:59] of the deal, AMD can buy Core Scientific [4:01] stock. And someone better call Dan [4:02] Toomey of Good Work because Lisa Su's [4:04] got the gong out. [4:06] >> We have a gong. We'd love you to hit it [4:08] to uh [4:09] >> on this [4:09] >> commemorate this day. [4:10] >> master day. [4:11] >> Do I get to do this? [4:12] >> Please. As hard as you want. [4:14] >> Gong. [4:15] >> There you go. Great hit. [4:16] >> AMD CEO hitting the gong and investing [4:18] in Anthropic seems fitting considering [4:20] that Anthropic CEO's wife previously [4:22] talked to Geoffrey Epstein about hitting [4:23] such gong while asking him for money. [4:25] But anyway, as for cousin Jensen Huang, [4:27] Nvidia reportedly backed up $250 billion [4:30] for OpenAI to lease a data center in [4:32] Ohio, which would be built by a company [4:34] owned by SoftBank, which is the same [4:35] Japanese investment firm that owns 13% [4:37] of OpenAI. And Jensen's BFF Howard [4:39] Lutnick gets the final say on whether or [4:41] not it'll be built. We have a clip of [4:43] Lutnick talking about his great [4:44] friendship. [4:45] >> And we had lunch on the island. That is [4:47] true. [4:48] >> Oops. Wrong friendship. That's a [4:51] That's a different one. [music] [4:52] >> You've been an incredible friend, an [4:54] incredible asset. And And as I said on [4:57] our first time we met, [4:58] >> [music] [4:58] >> you're a national treasure. [4:59] >> Nvidia's wheeling and dealing lately [5:01] doesn't stop with OpenAI though. It also [5:03] recently announced a half a trillion [5:05] dollar deal, $500 billion, [5:08] with South Korea's SK Group, which owns [5:11] SK Hynix. Nvidia also invested a billion [5:13] dollars in the South Korean company [5:15] Naver, gave a substantial pot of money [5:17] to the humbly named AI startup Safe [5:19] Superintelligence, and invested in a [5:21] startup called Volta that makes it [5:22] easier for startups to buy Nvidia chips. [5:24] Because if Nvidia can't give an AI [5:26] company money to buy its chips, it'll [5:27] pay someone else to do so. It really [5:29] does feel like Nvidia has become the [5:30] bank of Nvidia at this point. If you [5:32] look at things like the Huang [5:33] Foundation, Volta, and these other [5:35] efforts to move money around, it's a way [5:37] to keep the demand up on Nvidia's own [5:39] product, sourcing the money from Nvidia [5:41] to buy the product from itself or lease [5:43] it back or something like that with [5:45] other entities in between as extra [5:47] steps. [5:48] But, the big three memory manufacturers [5:50] are at play, too, because they continue [5:52] to signal to each other with their moves [5:54] in the market. SK Hynix and Samsung [5:56] joined Micron all in securing long-term [5:59] agreements to shore up memory pricing [6:01] and supply. SK Hynix finalized long-term [6:03] agreements, they say, with 10 customers. [6:05] And Samsung and Broadcom agreed to a [6:07] $200 billion deal for memory and foundry [6:10] capacity over 5 years. So, the memory [6:12] shortage driven by AI in the data center [6:14] buildout will continue to increase the [6:16] cost of consumer electronics and [6:18] decrease consumer access to technology. [6:20] Meanwhile, the federal government is [6:22] also busy. [6:24] Mostly gaslighting people. [6:27] To keep the AI boom booming. [6:29] >> You have some communities that really [6:31] want this. This isn't all negative [6:33] stuff. This is you have communities that [6:35] really want the data centers. [6:37] And frankly, those are the smart [6:38] communities because it means [6:41] a tremendous numbers of jobs, [6:43] uh very little actual disruption. [6:46] >> Now, despite the industry and the [6:47] government's best efforts, public [6:48] sentiment has turned against AI, but [6:51] OpenAI CEO Sam Altman knows how to fix [6:54] this. [6:55] >> I have been thinking about how we can [6:57] like organize field trips to a gigawatt [6:59] data center for people. [7:00] >> That, uh [7:02] that might be the most out of touch that [7:04] he's ever sounded. [7:05] >> I cannot imagine having [music] gone [7:07] through the like figuring out how to [7:08] raise a newborn without ChatGPT. [7:10] Clearly, people did it for a long time, [7:11] no problem. [7:12] >> Yes. [7:15] >> We brought you this video with [7:16] store.gamersnexus.net and our upcoming [7:19] actually human-crafted music videos, [7:20] documentaries, and movies on CD and [7:22] Blu-ray listed now on the store. These [7:25] include a special black market disk [7:27] version with a booklet containing a [7:28] written text adaptation available in [7:30] magazine-style print. We also have [7:32] another disk case with retro IO floppy [7:35] internal styling that we designed and 3D [7:37] print for each order, including a [7:38] spinning knob for mechanical [7:40] functionality. We're using this one for [7:42] our EVGA finale series educational [7:44] factory tour episodes and documentaries [7:46] like our Intel Fab tour. Finally, our [7:48] third version is an audio CD that [7:50] features all of the custom music we [7:52] hired Aaron Pauley from Of Mice and Men [7:54] for. When we asked him what genre to [7:56] categorize the mix of cyberpunk and [7:57] chiptune, he told us human-crafted. The [8:00] AI GPU black market and smuggling [8:01] documentary just turned 1 year old, so [8:03] we're launching a special disc version [8:04] of our Bloomberg upload. It features an [8:06] in-house and human-designed 3D printed [8:08] custom case, a Blu-ray disc with the [8:10] documentary on it, a second disc [8:12] included with bonus videos, and our [8:14] Bloomberg screenshot and peak irony, and [8:16] several of our disc collections also [8:18] include options with USB drives [8:19] containing high-quality MP4s that beat [8:22] our YouTube quality while still making [8:23] it physical. This is a great way to [8:25] support us directly and get physical [8:26] media in return. We're entering [8:28] production for these now. We'll be [8:29] burning the discs in-house and printing [8:30] the cases in-house over the next 3 [8:32] months, at which point we'll begin [8:33] shipping them. Head to [8:34] store.gamersnexus.net to learn more. [8:36] >> AI absurdity has hit new highs as the [8:38] industry scrambles to find something to [8:41] show for all of its efforts. AI grits [8:42] has swallowed the entire technology [8:44] industry with interconnected agreements, [8:46] deals, investments, and dependencies [8:47] where the same companies keep passing [8:48] dollar bills back and forth. And the AI [8:50] mania is led by OpenAI chief absurdist [8:53] Sam Altman and Nvidia kingmaker Jensen [8:56] Huang. [8:57] >> Somehow most people seem to still really [8:59] prefer interacting with a human. [9:01] >> Yes. [9:03] Somehow [9:05] >> It is made up that there's going to be a [9:07] singularity. [9:08] >> We are now like in the singularity. Like [9:10] this is the moment. [9:11] >> Don't create a narrative. Don't create a [9:12] story that is made up. [9:14] >> We are about to create [music] a genie [9:16] that can grant any wish. [9:17] >> And I encourage everybody to to [9:20] >> [music] [9:20] >> use cloud services as much as possible. [9:22] Everybody should use ChatGPT and [music] [9:24] Claude. [9:25] >> Well, I have a huge amount of context. I [9:26] mean, I'm like a AI model that's not [9:28] that smart. It's complete nonsense. [9:30] >> On July 26th, The Wall Street Journal [9:32] reported that Nvidia was in discussions [9:33] to provide a quarter trillion-dollar [9:35] backstop for OpenAI as part of a 10 [9:38] gigawatt data center project in Ohio. [9:40] The project may cost more than half a [9:42] trillion dollars and be the largest data [9:44] center announced to date. According to [9:47] the journal, open AI would lease the [9:48] data center project from SoftBank's [9:50] energy subsidiary. Nvidia's involvement [9:53] would enable the developers to quote [9:54] raise debt at more favorable terms than [9:57] it could if open AI had no financial [9:59] backer. Since open AI has no investment [10:02] grade credit rating as an unprofitable [10:05] private company. End quote. The Wall [10:06] Street Journal sources also said that [10:08] the US government controls the power for [10:10] the project with US Commerce Secretary [10:13] and Epstein Island tourist Howard [10:15] Lutnick involved in deciding which [10:17] companies would be involved. The deal [10:18] would be yet another of the circular [10:20] variety among Nvidia, open AI, and [10:22] SoftBank. SoftBank has invested more [10:25] than 64 billion dollars in open AI, [10:27] which translates to about 13% of the [10:29] company. And open AI invested 500 [10:32] million dollars into SoftBank's energy [10:34] subsidiary SB Energy last January. [10:36] Nvidia said it would invest 30 billion [10:39] dollars in open AI despite backing out [10:41] of its previous stance to invest up to [10:43] 100 billion dollars in the company. [10:45] However, open AI may actually need more [10:47] money if they want to put hardware in [10:49] the data center after it's built. The [10:51] journal reported Nvidia is discussing a [10:53] deal to finance the chips too, which [10:55] could run at 350 billion dollars. [10:57] Converted to a more stable currency, 350 [11:00] billion dollars is approximately equal [11:01] to two yuans [11:03] or about 1/3 of an Elon Musk. [11:06] No, wait. 1/5. No, no, wait. 1/2 now. [11:09] No, wait. CNBC confirmed the Wall Street [11:11] Journal's reporting about the 250 [11:13] billion dollar backstop. Nvidia declined [11:14] to comment to CNBC. We couldn't [11:16] independently verify the reports, but [11:19] they follow Nvidia's strategy to [11:20] financially support AI companies that [11:22] buy Nvidia's chips. Nvidia is still an [11:25] engineering company. It's just it seems [11:27] like it's more of a financial [11:28] engineering company now rather than [11:30] engineering engineering company. Where [11:32] in instances that it can't lock in [11:34] companies to its hardware with things [11:36] like CUDA, it instead moves to lock them [11:40] in or at least curry corporate favor [11:42] with [11:43] clever financial dealings. While Nvidia [11:46] has thrown its money all over the place [11:47] to invest in AI companies, [11:49] >> We don't pick winners. [11:51] And we we like we we we need to support [11:54] everyone. [11:55] And it's part of our part of our our our [11:57] joy of doing so, it's it's imperative [11:59] [music] to our business, but we also go [12:01] out of our way not to pick winners. And [12:02] so when I when I invest [music] in one [12:04] of them, I invest in all of them. [12:06] >> AMD hasn't invested in the AI ecosystem [12:08] in the same capacity. When Nvidia [12:10] offered to invest up to a hundred [12:12] billion dollars last year in OpenAI, AMD [12:14] countered by giving up a piece of itself [12:16] instead. AMD issued a warrant to OpenAI [12:19] that would allow OpenAI to buy up to a [12:20] hundred sixty million shares of AMD [12:22] stock that would vest over time based on [12:24] predetermined milestones. Jensen seems [12:27] perplexed at AMD's decision. [12:29] >> It's imaginative. [12:32] It's [12:33] it's unique and surprising. Considering [12:36] considering they were so excited about [12:38] their next generation product, I'm I'm [12:40] surprised that [12:41] they would give away [music] ten percent [12:42] of the company before they even built [12:44] it. And so [12:45] anyhow, it's it's uh [12:48] clever, I guess. [12:49] >> It's tough to live down disappointment [12:51] from family, but AMD is done playing [12:53] second fiddle to Nvidia. It wants in on [12:56] the bank rolling action for the AI [12:57] ecosystem as well. We've seen a trend [13:00] where AMD and Nvidia follow each other [13:03] with their investments. In May, AMD [13:05] revealed in an SEC filing that it had [13:07] purchased nearly six and a half million [13:09] dollars in Marvell Technology stock. [13:11] Marvell Technology makes custom silicon [13:13] products and brands itself as enabling [13:15] AI infrastructure. Two weeks later, [13:17] Nvidia announced a two billion dollar [13:19] investment in Marvell. Last November, [13:21] Nvidia and Microsoft announced that they [13:22] had invested up to ten billion dollars [13:24] and up to five billion dollars [13:25] respectively in Anthropic. On July 22nd [13:28] this year, AMD and Anthropic announced [13:30] an agreement wherein Anthropic would [13:31] deploy up to 2 gigawatts with AMD's data [13:33] center hardware. According to AMD's [13:35] press release, quote, Anthropic will [13:36] deploy AMD Helion rack-scale solutions [13:38] featuring AMD Instinct MI455X GPUs. Part [13:42] of the AMD Instinct MI450 series [13:44] together with AMD Epic Venice CPUs, AMD [13:46] Pensando networking, and ROCm software, [13:48] end quote. The deployment would build on [13:50] Anthropic's use of AMD's MI355X GPUs. [13:53] The first deployment would begin in [13:54] 2027. AMD and Anthropic would also [13:56] partner to optimize Claude, Anthropic's [13:59] family of language models, on AMD GPUs [14:01] and accelerate AMD's ROCm software. AMD [14:04] will adopt Claude internally with its [14:06] engineering and product development [14:07] teams. In return for all this, AMD said [14:09] it invest up to 5 billion dollars in [14:11] Anthropic. AMD's consumer strategy for a [14:14] long time now has been Nvidia GPU price [14:16] minus 50 dollars. And it seems like [14:18] they're applying that same strategy to [14:20] their AI bankrolling, where it's Nvidia [14:24] investment minus [14:26] 5 billion dollars. [14:28] The scale's a little different, but it's [14:29] kind of the same strategy. And AMD CEO [14:32] Lisa Su isn't bothered by sharing pieces [14:35] of the same companies with Nvidia, [14:37] because why would they be? [14:39] >> Is there any concern that everyone [14:41] in tech is so intertwined? [14:46] >> I think it's always been like that, [14:47] right? I mean, I think it's an ecosystem [14:49] where we're working on this together, [14:51] and in some sense, [14:53] I think we are uh probably more aligned [14:57] than we've ever been, uh which is [14:58] actually a really good thing. When I say [15:00] aligned, it's because these are such [15:03] large investments over time, the more we [15:06] know about um what our customers need, [15:09] the better prepared we we can be for uh [15:12] for that moment. So, I think that's what [15:15] um the the AI world is such that um you [15:19] need every part of the ecosystem to come [15:21] together, and and so the fact that the [15:24] ecosystem is intertwined is the way it [15:26] should be. [15:27] >> Everybody is so friendly. Everyone loves [15:29] each other. [15:31] Even if they're competitors, it doesn't [15:33] matter. [15:33] >> But despite all being friends or [15:36] something even when they're competing, [15:38] it's okay because they occasionally keep [15:41] the facade up. AMD made a competitive [15:43] jab at Nvidia on July 28th when it [15:45] announced a new infrastructure agreement [15:47] with the Neo Cloud Core Scientific. As [15:49] part of this agreement, AMD would secure [15:51] quote up to 2 and 1/2 gigawatts of data [15:53] center capacity to support end customer [15:55] deployments of AMD AI solutions, end [15:57] quote. AMD's AI ecosystem would increase [15:59] by 500 megawatts starting in 2027. Core [16:01] Scientific issued AMD a warrant that [16:03] would allow it to buy up to 30 million [16:05] shares of Core Scientific stock for [16:07] $23.47 per share. The shares would vest [16:10] over time as Core Scientific deploys AMD [16:12] hardware in its data centers. The whole [16:14] thing's pretty interesting because the [16:15] deal shows the budding interest in Neo [16:17] Clouds or companies that provide GPUs as [16:20] a service in data centers. [16:22] Uh and AMD's desire to own stock [music] [16:26] in companies that buy AMD's GPUs. It [16:29] really does mirror a lot of what [16:31] Nvidia's doing. Core Scientific began as [16:33] a crypto mining company before pivoting [16:35] to AI. Out of the grifting pan and into [16:37] the grifter, as they say. The company [16:39] had a tumultuous first few years as a [16:41] public company. Core Scientific was [16:43] listed on the Nasdaq in 2022, unlisted [16:45] in 2023, and then listed again in 2024. [16:47] This is a good sign. Nvidia sugar Weave [16:50] attempted to acquire Core Scientific in [16:51] 2024 and 2025 but ultimately never [16:54] succeeded. The Neo Cloud industry rents [16:56] out GPU capacity, making it a keystone [16:58] at this point in this entire industry [17:00] where they've helped to manage make [17:03] consumers obsolete [17:05] because they just buy the GPU capacity [17:07] instead and then they rent it to their [17:08] customers. AMD doesn't want Nvidia to [17:11] completely own the Neo Cloud business [17:14] because that seems to be where they [17:16] think the business is going. Nvidia has [17:18] invested in many such neo cloud [17:20] companies including Coreweave, Nebius, [17:22] and Scale and Iren. Coreweave in [17:25] particular, as we've often not joked but [17:28] stated very factually, is Nvidia's sugar [17:30] baby. [17:31] >> If we didn't help Coreweave exist, they [17:33] would not exist. [17:34] >> Coreweave exclusively uses Nvidia GPUs [17:37] and Nvidia owns more than $3 billion of [17:39] Coreweave stock or about 11% of the [17:40] company. As we previously reported, even [17:43] Jensen Huang's non-profit foundation [17:45] gave Coreweave $108 million for [17:47] computing capacity. So much for [17:48] non-profit when it all just kind of [17:50] circles back to you anyway, back to the [17:52] origin whence it came. AMD though has [17:55] been facing pressure from Nvidia to make [17:58] similar such investments and this we [18:00] think could potentially jeopardize AMD [18:03] disproportionately from Nvidia if [18:05] everything were to all suddenly [18:07] collapse. AMD and Nvidia also both [18:09] previously invested in Elon Musk's AI [18:11] company, although that isn't working out [18:13] for AMD. See now as Elon has called [18:16] Nvidia GPUs, quote, "the best," end [18:19] quote. The AI industry has become [18:21] interlocked and dependent on each other. [18:23] We've known this for quite some time now [18:25] as they continue to expand the base of [18:27] the triangular economy that they're [18:29] building around all this and lately the [18:31] move over the past several months has [18:33] been to fortify the AI bubble we think [18:36] with investment dollars that ultimately [18:38] kind of go into data centers and neo [18:40] clouds [18:41] while they rake in billions of dollars [18:44] in revenue. And case in point, you can [18:47] look at the memory manufacturers where [18:49] there's still just three of them [18:50] controlling most of the market. CXMT is [18:52] a different conversation. We have a [18:53] documentary on them. That's the Chinese [18:56] memory manufacturer that's been on the [18:57] rise lately. But Samsung, SK Hynix, and [19:00] Micron as the big three have maintained [19:02] a unified message. No surprise that [19:05] there's no end in sight to the memory [19:07] shortage. All three memory companies are [19:09] saying the the thing [19:11] just like they've done multiple times in [19:13] the past when being accused of and at [19:15] least one instance convicted of being a [19:17] cartel. [19:19] Uh but we'll start with Samsung for the [19:20] modern stuff. Samsung reported record [19:23] earnings in the second quarter of 2026 [19:25] with a 28% quarter-over-quarter increase [19:27] in revenue. Samsung's memory business [19:29] set a new record for quarterly revenue [19:31] and operating profit by, quote, [19:33] "proactively addressing AI demand [19:35] despite limited capacity with a primary [19:37] focus on server products." "The [19:38] continued industry-wide upward trend of [19:40] prices also contributed to the record [19:42] earnings." End quote. Further, Samsung [19:44] expects continued growth for server [19:46] DRAM, SSDs, and HBM that'll keep the [19:48] market, quote, "undersupplied despite [19:51] partial demand moderation in mobile and [19:54] PCs." End quote. Partial demand [19:57] moderation is an impressively creative [20:01] euphemism for uh [20:03] nobody can afford memory. Samsung also [20:05] said these, quote, "supply constraints [20:07] are expected to continue." End quote in [20:08] its press release and reiterated this in [20:10] an earnings call. The Register quoted [20:12] Samsung's memory EVP as saying, quote, [20:14] "The price increases will continue until [20:16] morale improves." [20:19] Oh, that No, wait. That doesn't [20:20] Oh, it it doesn't actually say that. I [20:22] misread that. Quote, "Based on the [20:24] incoming request that we have been [20:25] seeing from the customers, unmet demand [20:27] from this year is likely to carry over [20:29] into the following year, contributing to [20:31] the tighter supply conditions going [20:33] forward. The supply constraints are [20:34] expected to become even more severe in [20:35] 2027 than 2026, reinforcing our view [20:38] that the supply shortage will persist [20:40] through 2028." End quote. These [20:42] shortages are good for Samsung's [20:44] coffers. As shown in an earnings [20:45] presentation, Samsung's semiconductor [20:48] business sales grew by 56% [20:50] quarter-over-quarter. Samsung recorded [20:52] 89 and 1/2 trillion Korean won in total [20:54] operating profit with almost all of [20:57] that, 89.2 trillion won, coming from the [21:00] semiconductor business. That's about $62 [21:03] billion [21:04] in US currency in one quarter. South [21:06] Korea's other major memory manufacturer, [21:08] SK Hynix, is making similar predictions [21:11] about the memory shortage. In an [21:12] interview with Reuters from July 10th, [21:14] the SK Hynix CEO said, quote, "We [21:16] forecast that next year will be the [21:18] worst year in the industry's history [21:20] from the supply perspective. We still [21:22] forecast that customer demand will [21:23] remain higher than our supply capacity [21:26] even beyond 2030, but we are doing our [21:28] best to solve the problem." End quote. [21:30] At the end of July, SK Hynix reported [21:32] its second quarter results showing that [21:34] SK Hynix generated 79.3 trillion won in [21:37] revenue and 60.5 trillion won in [21:40] operating profit. Its operating margin [21:42] was 76%. SK Hynix said that, quote, [21:45] "Revenue and operating profit increased [21:47] by 257% [21:50] and 557% year-over-year, respectively." [21:53] End quote. SK Hynix also secured, they [21:56] said, around 10 long-term agreements [21:58] with key customers. In an earnings [22:00] presentation, Hynix said that these [22:02] agreements provide Hynix with, quote, [22:04] "Pricing structures designed to address [22:06] price volatility." End quote. These [22:08] agreements would allow SK Hynix to [22:10] better predict demand and smooth out the [22:12] memory market's boom and bust cycles. [22:14] Stated another way, it also helps them [22:16] keep the prices elevated by locking them [22:18] in with key clients. This is similar to [22:20] the approach that Micron took to lock in [22:22] price. In its last earnings, Micron [22:25] announced 16 strategic customer [22:27] agreements with the largest agreements [22:29] giving customers a price ceiling and a [22:31] price floor. To no one's surprise, [22:33] Micron also said the memory shortage [22:35] will persist through 2027. And who could [22:38] have guessed that the big three are [22:40] all saying and doing precisely the same [22:44] thing. [22:45] Again. [snorts] [22:46] Again. [22:48] But, there's one key difference. [22:50] Micron CEO [22:52] knows who to blame for the crunch. [22:56] It's it's it's the customers. [22:58] It's the customers' fault. [22:59] >> Then we were coming out of the 2023 time [23:03] frame, uh which is when actually the [23:05] industry, certain customers drove [23:08] pricing significantly down in our [23:10] industry, and that of course resulted in [23:13] significant negative gross margins, you [23:16] know, for certain customers, certain [23:18] products, and that really impacted the [23:22] investment capability of the industry. [23:24] >> If only Micron had more tax breaks and [23:27] tax incentives and free government money [23:30] to build fabs, then it could resolve the [23:33] shortage problem. [23:35] I mean, not for you or me, but, you [23:37] know, like for Nvidia and stuff. The [23:39] memory shortage isn't going away soon. [23:41] Micron, SK Hynix, and Samsung continue [23:44] to set quarterly revenue records, and [23:46] all three predict the memory shortage is [23:48] here to stay, at least for a while. When [23:50] the memory market is tight, all three of [23:52] these companies do well. It's always [23:54] been the case for them. [23:56] But, enter a new player, federal [23:58] gaslighting. Despite community after [24:01] community blocking data center [24:02] development or establishing moratoriums, [24:04] the federal government will do whatever [24:06] it can to keep the good times rolling [24:08] for AI companies. President of the [24:10] United States and Nvidia shareholder [24:12] Donald Trump, who has traded millions of [24:15] dollars of Nvidia shares while [24:16] president, really wants you to like data [24:19] centers. [24:19] >> You have some communities that really [24:21] want this. This isn't all negative [24:23] stuff. This is you have communities that [24:25] really want the data centers. Very [24:27] important because [24:28] you have to convince your community how [24:30] great these things you can't fight it. [24:32] You have to go with it, and you're going [24:34] to have all that protection, and we're [24:35] giving you automatic protection, but [24:37] you're going to make a tremendous amount [24:39] of money, you're going to reduce taxes, [24:41] and you're going to have something that [24:42] if you don't have it, everybody else [24:44] wants it. So, uh [24:47] these are the smartest people. [24:49] >> That's right. Trump wants you to not [24:50] fight it, to just go with it, and to [24:52] trust the protection. All things that [24:55] definitely [24:57] don't sound really creepy and [24:59] concerning. Trump made these remarks [25:01] during a meeting about the ratepayer's [25:03] protection pledge, which is the [25:05] government's claimed attempts to keep [25:06] data centers from increasing electricity [25:08] bills for Americans. As part of the [25:10] pledge, the government asks companies to [25:12] quote build, bring, or buy and quote [25:16] energy resources to pay for their data [25:18] centers electricity costs. [25:20] BYO energy. More than 280 organizations [25:24] have signed the pledge to date, [25:25] including Microsoft, Open AI, Google, [25:28] Meta, Oracle, and Coreweave. However, [25:30] the pledge has also been criticized for [25:32] its lack of enforcement mechanisms. It's [25:34] also voluntary. And Trump even called [25:37] this earlier in the year good PR for [25:39] data centers. And that's really all it [25:41] seems to be because one of the obvious [25:43] concerns with this is deregulation. [25:46] Power is not really something you want [25:48] to with without good regulation and [25:51] understanding from everybody as to what [25:54] it is the power generating utility is [25:58] that's going near their towns. And in [26:01] this case, forcing companies to bring [26:02] power internal leads to, we think, [26:05] dangerous build-outs with minimal [26:07] oversight and sets us back decades in [26:10] terms of things as simple as is the air [26:12] you're breathing clean and safe to [26:14] breathe. You see this, for example, like [26:17] with Musk's Memphis facility in [26:19] Tennessee that's been challenged because [26:21] it is running gas-operated turbines [26:24] power on top of the facility which [26:27] residents contest has caused them to [26:29] develop breathing conditions and [26:31] disorders. [26:32] A couple days after the ratepayer [26:34] protection pledge briefing, the [26:37] Environmental Protection Agency decided [26:39] to [26:40] you'll never believe this. [26:43] Not protect the environment. [26:46] Didn't They They didn't do that part. On [26:49] July 27th, the Environmental Protection [26:51] Agency clarified a rule wherein data [26:53] centers that use on-site power [26:55] generation facilities that aren't [26:57] connected to a public grid are not [26:59] subject to regulations within the Clean [27:01] Air Act acid rain program. Yeah, just a [27:05] just uh [27:07] if I could just get the government for a [27:08] second, um [27:10] when you're talking about good PR from [27:12] an energy program, normally you don't [27:15] want it to involve sentences such as [27:18] not being subject to [27:20] the acid rain program. [27:23] Just from a PR standpoint sounds kind of [27:25] bad. I'm sure it's [27:27] >> [music] [27:28] >> not [27:29] though, right? The EPA refers to these [27:32] as islanded power generators. In a [27:34] release, the EPA said, quote, "The [27:36] guidance expands opportunities for [27:38] companies to develop and operate [27:40] islanded power generation facilities for [27:42] data centers." End quote. [27:43] >> EPA! EPA! [27:45] EPA! [27:46] >> [screaming] [27:47] >> The acid rain program is part of the [27:49] Clean Air Act and it would require [27:51] emissions reductions of sulfur dioxide [27:53] and nitrogen oxides, two chemicals that [27:55] cause acid rain, from power facilities. [27:58] Even the EPA has hailed the program as a [28:01] success for reducing fossil fuel [28:03] emissions and benefiting the environment [28:05] and human health. But that's the key. [28:07] Human health. [28:09] We don't need those where we're going. [28:11] The EPA is helmed by Lee Zeldin, who [28:15] during his Senate confirmation hearing [28:18] would not state that the US should [28:20] reduce its reliance on fossil fuels. [28:23] Likewise, in his Senate confirmation [28:26] hearing, he wouldn't state clearly, [28:28] would not accept that the EPA has a, [28:31] quote, obligation, end quote, to protect [28:33] Americans from disaster accelerated by [28:36] climate change. The EPA's rule changes [28:38] also fall in line with the government's [28:40] AI action plan, which Lisa Su has been [28:43] an advocate of. Last year, the federal [28:45] government said it would reduce rules [28:47] related to the Clean Air Act and Clean [28:49] Water Act to expedite permitting for [28:52] data center development. But the EPA's [28:54] limp-wristed responses and reactions [28:56] align with the federal government's [28:58] expansion of its balance sheet. Last [29:00] year, the government took a percentage [29:03] of Intel, a stake in the company, and [29:06] likewise, the government wants a piece [29:08] of yet more semiconductor companies. On [29:10] July 29th, the Department of Commerce [29:13] signed letters of intent with seven [29:15] semiconductor companies to provide $874 [29:18] million as part of the CHIPS and Science [29:20] Act, which the president has bashed [29:22] relentlessly despite both his [29:24] administration and Biden's [29:26] administration funding the same act. [29:29] >> We had this ridiculous CHIPS Act where [29:31] you give billions of dollars to a [29:32] company and they didn't even know what [29:33] the hell to do with it. [29:35] But this was the Biden fools, those [29:37] idiots that almost destroyed our [29:39] country. [29:40] But we're also reasserting the [29:42] fundamental truth to be a strong nation, [29:45] you must be a manufacturing nation. [29:47] There are manufacturing jobs. [29:49] >> The federal agencies said that the $874 [29:51] million would, quote, support innovative [29:54] domestic technologies to dramatically [29:56] increase the performance of the world's [29:57] fastest computers, secure domestic [29:59] supply chains, and strengthen US [30:01] leadership in the compute supply chain, [30:03] end quote. The Commerce Department would [30:05] also receive a, quote, minority [30:07] non-controlling equity stake in each [30:10] company as a condition for receiving the [30:12] funds to enhance the return for the US [30:14] taxpayer, end quote. So once again, the [30:17] capitalism isn't capitalizing. [30:21] It's a little different strategy. [30:23] Semiconductor manufacturer [30:24] GlobalFoundries, which got stranded on [30:27] 14 nanometer for years, received the [30:29] most funding at $300 million. [30:32] GlobalFoundries would use the money for [30:34] research and development for, quote, [30:36] next-generation optical materials, wafer [30:38] technologies, and advanced packaging, [30:40] end quote. GlobalFoundries said the [30:42] government would take about 1% equity in [30:44] the company as part of a separate [30:46] agreement. So, the tech industry is not [30:48] just in an AI bubble, but in AI [30:50] psychosis as it looks towards clowns who [30:54] are CEOs of these companies [30:57] and are officials in government for some [31:00] form of guidance [31:02] or in the case of Lisa Su, some kind of [31:05] weird approval seeking. [31:06] >> So, what do you think about that? Is [31:07] that a big number? [31:08] >> Nvidia has effectively turned into a [31:10] bank for AI companies so that data [31:12] centers will be filled with Nvidia [31:13] hardware. Nvidia and AMD are investing [31:16] billions of dollars in the same AI [31:18] companies. Micron, Samsung, and Hynix [31:20] keep breaking earnings records while [31:22] downplaying the impacts on consumers and [31:25] multiplying their margin on consumer [31:27] products. The federal government wants [31:29] to accelerate data center construction [31:31] at the expense of the environment and [31:33] your health. All the while the tech CEOs [31:36] of these companies keep making [31:37] hyperbolic statements about what AI will [31:40] or won't do. Every week it's some new [31:43] story about how some AI broke out of [31:46] containment from Anthropic or OpenAI or [31:48] whoever it may be and it's this big [31:50] scary thing that can just hack anyone [31:53] without any effort whatsoever. As if [31:55] these companies don't have the [31:56] technological capability to do a basic [31:59] IT admin level job of not giving it [32:03] access to the internet. No, no, this is [32:05] legitimate. It can definitely just [32:08] breach through all these different [32:10] layers [32:11] and get out onto the open web and start [32:13] wreaking havoc. Surely, this is not some [32:15] kind of PR campaign to boost the [32:18] perceived value of the product that they [32:20] pedal before they move towards IPOs. [32:23] That's definitely not it. That would Why [32:25] would they do that? [32:27] >> I like money. [32:28] >> OpenAI CEO Sam Altman said AGI has [32:31] arrived. Nvidia CEO Jensen Huang said [32:33] the global GDP will increase to $500 [32:35] trillion. dollars. SpaceX CEO Elon Musk, [32:38] said working will be optional in the [32:40] future and that money will be [32:41] irrelevant. Anthropic CEO, Dario, said [32:44] AI could increase unemployment up to 10 [32:46] to 20% within the next 5 years. AMD CEO, [32:49] Lisa Su, said AI can be your friend. [32:52] >> The more we use it, the more it can help [32:55] you. And this is all about getting [32:57] consumers familiar with the technology [32:59] and trusting that the technology can [33:01] really be like your best friend. [33:03] >> And so we come back to the quote from [33:05] Altman at the start of this video. [33:07] >> People seem to hate these things so [33:08] much. What's to be done about this? [33:10] >> First of all, I have been thinking about [33:12] how we can like organize field trips to [33:14] a gigawatt data center for people [33:16] because it is one thing to say [33:19] it is another thing to see a photo or a [33:20] video of and then it's a whole other [33:21] thing to just stand in person there and [33:23] be like, "Oh man, [33:26] this is like an unbelievable scale. [33:28] Building one of these is like order of [33:29] 10,000 construction workers going [33:31] full-time for a year and a half. [33:33] The energy that flows through one of [33:35] these things could power a small city. [33:36] Again, we've just like lost all sense of [33:38] scale, but these would have been among [33:40] each of these would have been among the [33:42] most in expensive infrastructure [33:44] projects the team has ever done and now [33:46] we've done a lot of them. [33:47] >> AI chip maker, Jensen Huang, used to [33:50] vehemently deny the existence of a [33:52] bubble in AI. Actually, Lisa Su has done [33:55] the same. But there was an interesting [33:57] change recently where in a recent [33:59] interview with Axios, Jensen Huang's [34:01] narrative took a turn. And when asked a [34:04] question about the dangers of an AI [34:06] bubble, he said this. [34:07] >> The bubble will come someday. It's just [34:09] not today. [34:10] >> Saying the bubble will come someday is a [34:13] big difference from denying its [34:14] existence altogether. And when die-hard [34:18] GN fan, Bloomberg, asked Jensen about it [34:21] last year in November, he said this. [34:24] >> Was that you saying we are not in an AI [34:27] bubble? [34:28] >> I don't believe we're in an AI bubble. [34:30] And the reason for that is [34:31] we're going through a natural transition [34:34] from a old computing model based on [34:36] general purpose computing to accelerated [34:38] computing. We also know that AI has now [34:41] become good enough because of reasoning [34:43] capability, research capabilities, its [34:45] ability to think, it's now generating [34:47] tokens and now generating intelligence [34:50] that's worth paying for. [34:51] >> So, there's a common thread here. The [34:53] same executives who keep making lofty [34:55] predictions about how powerful and scary [34:58] and important AI is are also the ones [35:01] who are getting filthy rich. [35:04] As they rake in through their companies [35:07] uh all kinds of tax breaks, incentives, [35:12] free money, investment from the [35:13] government, government-aided bailouts of [35:16] assets. However, companies are beginning [35:18] to question the ROI of their AI [35:20] investments. Investors appear to be more [35:22] wary, and everyday people are dealing [35:24] with the consequences from higher prices [35:26] for consumer goods to increased air [35:28] pollution and rising electricity bills. [35:31] The American public overwhelmingly [35:32] disapproved of data centers being built [35:34] near them, and the public is rallying [35:36] against them in their communities. [35:38] According to research firm Data Center [35:40] Watch, local communities blocked or [35:42] delayed 75 data center projects worth [35:44] $130 billion in the first quarter of [35:47] 2026. Most people don't want data [35:49] centers near them. They don't want to [35:51] hear them or see them. And unless [35:54] they're Jim Cramer, [35:56] they don't want to live in them. [35:57] >> I like data centers so much. I WISH I [36:00] LIVED IN ONE. [36:01] >> THAT'S complete nonsense. [36:02] >> Maybe it's time those two get a room. [36:05] Preferably in a data center. [36:08] Away from everyone else. [36:10] Ideally in the section that's [36:11] underground. [36:13] That's it for this one. Thanks for [36:14] watching. Subscribe for more. You can go [36:17] to store.gamersnexus.net to support us [36:19] directly in our efforts or [36:21] patreon.com/gamersnexus [36:23] if you want to throw a few bucks our way [36:24] each month. All this helps us with our [36:27] deep dive reporting, which can take a [36:29] long time to put all this research [36:30] together and [36:32] and then bring it to you all through [36:34] video. So, thanks for watching. We'll [36:36] have plenty more to talk about. [36:38] Subscribe for all that and we'll see you [36:40] all next time.