DRW's Wilson Bets GPU Futures Will Cut Neocloud Capital Costs
Bloomberg
Standardized compute futures will lower borrowing costs for neoclouds by allowing operators to hedge hardware depreciation through synthetic basis trades.
AI-assisted summary of Don Wilson on the GPU Market's Trillion Dollar Question — original source. Check the original for full context.
Source published:
Key ideas
- DRW and Silicon Data are targeting the launch of cash-settled GPU futures contracts on the CME, potentially eligible for listing following the expiration of the CFTC comment period on November 9th. (5:14)
- Spiking bond yields up to 15% for neocloud provider CoreWeave highlight severe financing bottlenecks that could be eased by using forward GPU strips as synthetic basis hedges against hardware capex. (3:47)
- DRW is developing the 'crunch spread,' a derivative concept tracking the processing margin between raw GPU rental costs and output token market prices, modeled after oil crack spreads and soybean crush spreads. (15:00)
- Don Wilson advocates for perpetual futures in traditional finance, arguing they eliminate roll costs for retail commodity ETFs and democratize borrow fees on hard-to-borrow equities. (26:15)
Detailed notes
## Background Published by Bloomberg on October 19, 2026, as an episode of the Odd Lots podcast, this live recording from City Winery in Chicago features hosts Joe Weisenthal and Tracy Alloway interviewing DRW founder and CEO Don Wilson. The discussion revisits Wilson's vision of establishing a financial compute market bigger than oil, examining recent AI tech jitters, neocloud credit risks, and the emergence of GPU and token derivatives.
## TL;DR Don Wilson addresses recent anxieties in the artificial intelligence sector, dismissing growth volatility and CoreWeave credit concerns as routine growing pains of an expanding hypergrowth market. He outlines DRW's initiatives to standardize compute derivatives, explaining how forward GPU futures curves can lower capital costs for neoclouds via synthetic basis trades. Wilson introduces the concept of the 'crunch spread'—the margin between raw GPU compute costs and generated token outputs—while detailing the economic merits of benchmark hardware contracts. Beyond compute, he defends perpetual futures against regulatory skepticism, advocating for their utility in commodity ETFs and hard-to-borrow equities while emphasizing the inevitability of 24/7 global trading.
## Atmosphere & style The live recording at City Winery carries an energetic, intellectually sharp, and collegial atmosphere, pairing relaxed stage banter with sophisticated financial discourse. Don Wilson projects calm, pragmatic confidence, easily cutting through retail tech panic with the seasoned perspective of a veteran derivatives trader. The hosts sustain a dynamic, fast-paced rhythm, playfully drawing attention to industry luminaries in the audience like CME Chairman Terry Duffy and turning dry regulatory debates over cleared swaps into engaging, thought-provoking dialogue.
## Bulletin - `05:14` DRW and Silicon Data are targeting the launch of cash-settled GPU futures contracts on the CME, potentially eligible for listing following the expiration of the CFTC comment period on November 9th. - `03:47` Spiking bond yields up to 15% for neocloud provider CoreWeave highlight severe financing bottlenecks that could be eased by using forward GPU strips as synthetic basis hedges against hardware capex. - `15:00` DRW is developing the 'crunch spread,' a derivative concept tracking the processing margin between raw GPU rental costs and output token market prices, modeled after oil crack spreads and soybean crush spreads. - `26:15` Don Wilson advocates for perpetual futures in traditional finance, arguing they eliminate roll costs for retail commodity ETFs and democratize borrow fees on hard-to-borrow equities.
## Key takeaways - `00:00` Market anxiety surrounding AI revenue growth and neocloud credit spreads represents routine variance in hypergrowth technologies rather than structural decay. - `03:47` Neoclouds face high costs of capital when catering to lower-tier developers; selling forward GPU futures strips allows them to structure a basis trade that locks in baseline revenues and lowers borrowing rates. - `05:14` CME-targeted GPU futures could clear regulatory review as early as November 9th following an extended CFTC public comment period. - `05:14` Reliable cash-settled compute futures require transparent index methodologies complying with 2013 IOSCO benchmark principles, drawing pricing primarily from independent neocloud boards rather than opaque hyperscaler contracts. - `06:26` The long-term terminal value and depreciation rate of advanced hardware architectures like Nvidia's B200 represent the critical financial unknown driving the necessity for a 3-year forward curve. - `10:00` Compute pricing does not follow a simple decay curve; demand shocks from model releases can cause GPU rental prices to surge while token prices collapse simultaneously. - `10:00` Futures markets will not inherently subsidize reckless AI capex because rational forward curves adjust downward under heavy supplier selling pressure to attract speculative buyers. - `15:00` A diversified inference ecosystem powered by open-source hosting providers like Fireworks AI and Together AI provides sufficient breadth to support exchange-traded compute contracts. - `15:00` Nvidia H100 and B200 benchmarks will anchor compute derivatives, with specialized ASICs, custom chips, and networked training clusters priced as basis spreads relative to those core contracts. - `15:00` The 'crunch spread' formalizes the economic relationship between GPU compute rental inputs and model token outputs, paralleling crack and crush spreads in energy and agriculture. - `20:00` Multiple distinct contracts will be needed for dominant chips, following the precedent of power markets which trade thousands of localized nodal contracts. - `20:00` Nvidia is projected to expand its aggregate revenue and enterprise value even as alternative hardware options gradually erode its percentage market share. - `25:00` To manage exponential enterprise token consumption, trading firms must optimize costs by dynamically routing lower-complexity inference requests away from costly frontier models. - `26:15` Perpetual futures can solve the front-running and roll-cost drag inherent to dated commodity ETFs by continuously tracking underlying benchmarks without contract expiries. - `26:15` Holding perpetual equity contracts mechanically passes lucrative borrow fees directly through to long holders via funding rates, unlike traditional retail stock ownership. - `30:00` A centrally cleared swap and a futures contract carry identical societal counterparty risk, making legal debates over nomenclature largely obsolete. - `30:00` Excessive leverage and auto-deleveraging are exchange-level design choices rather than inherent economic characteristics of the perpetual contract structure. - `31:30` The continuous operations model established by cryptocurrency trading since 2014 will inevitably expand into traditional financial derivatives like SOFR options.
## Detailed walkthrough ### `00:00` Introduction: Live at City Winery Chicago and the AI Market Jitters
Recording live for the second consecutive year at City Winery in Chicago, hosts Joe Weisenthal and Tracy Allaway welcome Don Wilson, the founder and CEO of principal trading firm DRW. While Wilson appeared on the podcast roughly a year earlier to outline his vision for building a financial compute market, the hosts note that twelve months in the artificial intelligence landscape feels like an entire decade. Setting the stage for the discussion, Weisenthal points to recent jitters across the AI ecosystem: mounting credit concerns around neocloud provider CoreWeave, widening spreads on CoreWeave bonds, and market anxiety following an Anthropic investor meeting ahead of a potential IPO where revenue growth numbers were closely scrutinized. Asked whether the sector is experiencing a genuine slowdown or shifting into a lower gear, Wilson dismisses the panic as routine variance inherent to hypergrowth technologies.
Wilson explains that uneven progress is mathematically inevitable when a market expands at such breakneck speed, arguing that neither the first nor second derivative of growth can remain linear:
> “Any new thing that's growing this fast is not going to grow at a constant rate. It's not going to grow at a, you know, the first derivative isn't going to be constant. The second derivative isn't going to be constant. So, to me, this is just kind of expected volatility.”
He playfully speculates that end users may simply have taken vacations in August rather than burning through Anthropic tokens, reiterating that he sees nothing fundamentally broken in the broader trajectory.
---
`03:47` CoreWeave, Neocloud Credit Risk, and How GPU Futures Unlock Capital
Allaway turns the conversation toward CoreWeave and DRW’s initiatives—specifically Compute Exchange and Silicon Data—which aim to build standardized compute markets so cloud infrastructure providers can hedge future cash flows and lower their financing costs. With CoreWeave’s bond yields spiking to 15%, Allaway asks whether having liquid compute futures in place would resolve these structural bottlenecks. Wilson responds that the bond market's reaction underscores an urgent demand for formal risk-management instruments across the neocloud sector.
```mermaid flowchart TD accTitle: Neocloud Hedging and Basis Trade Mechanism accDescr: Diagram illustrating how a neocloud uses GPU futures to hedge capacity and lower its cost of capital.
subgraph Physical ["Physical Asset Deployment"] A["Neocloud buys GPUs<br/>(e.g., H100s, B200s)"] B["Deploy hardware in data centers"] A --> B end
subgraph Financial ["Futures Market Hedge"] C["Sell strip of GPU futures<br/>(e.g., CME 3-year curve)"] D["Lock in guaranteed baseline revenue"] C --> D end
subgraph Revenue ["Commercial Offtake & Financing"] E["Sell on-demand compute OR<br/>6-month contracts with upfront collateral"] F["Giant basis trade profile:<br/>Drastically reduced cost of capital"] E --> F end
B --> C D --> E ```
Wilson breaks down the fundamental divide in cloud counterparty risk: - **Tier-1 counterparties:** Neoclouds that sign multi-year capacity agreements exclusively with mega-cap, high-credit-rating companies face minimal counterparty pressure. - **Lower-tier & on-demand users:** A vast pool of AI startups and developers lack prime balance sheets, forcing neoclouds that cater to them to absorb heavy credit discounts and pay double-digit debt yields. - **The synthetic hedge:** By deploying physical GPUs and immediately selling forward strips of GPU futures, neoclouds create a synthetic basis trade. They lock in hardware economics while selling short-term compute or six-month blocks backed by upfront margin deposits.
Highlighting the transformative effect this financial plumbing would have on corporate balance sheets, Wilson emphasizes the borrowing advantage:
> “And the market is going to reward companies that do that, that basically just have on a giant basis trade, heavily, their cost of capital is going to be much lower.”
---
`05:14` Regulatory Hesitancy, the CFTC Comment Period, and IOSCO Index Robustness
Weisenthal asks why GPU futures do not yet exist, citing public reports of regulatory scrutiny and purported White House national security concerns surrounding the formal commoditization of AI chips. Wilson explains that the regulatory process remains open: the Commodity Futures Trading Commission (CFTC) extended its public comment period, and if the agency takes no formal action to block the filing, the contracts could become eligible for listing as early as November 9th. Wilson wryly observes that November 9th falls neatly after Election Day, suggesting political sensitivities in Washington may ease once votes are cast, though he admits he cannot fully decipher the exact policy objections.
The discussion pivots to whether underlying compute indices are reliable enough for cash-settled futures contracts. Because substantial GPU volume is transacted via confidential, bespoke bilateral offtake contracts between model developers and cloud operators, pricing can appear opaque. Wilson acknowledges that index construction varies widely across the industry:
> “Listen, I think it's buyer beware, learn about the index that you're trading futures on.”
He contrasts the market approaches: - **Hyperscalers vs. Neoclouds:** While hyperscalers dominate mega-deals shrouded in non-disclosure agreements, there are dozens of independent neocloud providers operating transparent on-demand pricing boards. - **Silicon Data’s methodology:** Silicon Data gathers broad on-demand pricing feeds exclusively from neoclouds to construct an index engineered to comply with IOSCO (International Organization of Securities Commissions) standards. - **Regulatory rigor:** Wilson remarks that while Silicon Data has targeted full compliance with the 2013 IOSCO benchmark principles for financial indices, competing data vendors often operate without even knowing what IOSCO compliance requires.
---
`06:26` The Trillion-Dollar Question: GPU Depreciation, the B200, and Price Volatility
Allaway confronts the central bear thesis against GPU futures: while an investor might be structurally bullish on long-term compute consumption, individual microchips are rapidly depreciating assets that face obsolescence as faster, more power-efficient architectures debut. Wilson agrees that valuing future processing power is the single most important puzzle in tech finance today.
> “That is the trillion dollar question. That is why these futures need to exist, because nobody knows what the value of a B200 will be four years from now, five years from now.”
Wilson reveals that the initial contracts slated for the CME are designed to extend three years down the forward curve, with longer maturities added as liquidity matures. He recounts how skeptics voiced the same doubts late last year when hourly rental rates for Nvidia H100 chips were steadily cratering: market participants questioned the utility of GPU futures, assuming chip rental prices would drift downward forever. However, around Thanksgiving, on-demand compute rates abruptly reversed course and surged higher, demonstrating that the market experiences aggressive supply-and-demand squeezes rather than a straightforward, predictable decay curve.
`10:00` Volatility, Depreciation, and the Economics of Forward Hedging
Recent market dynamics highlight the extreme volatility of compute pricing. Following the release of Anthropic's newest model—which marked a massive step up in usefulness—demand for inference surged almost overnight. Prices for hardware like Nvidia's B200, B300, and H100 saw a sharp upward spike, even as the Silicon Data token index price plummeted over the exact same window. As Don Wilson points out:
> “I mean, these things are volatile. They can go up, they can go down, and a hedging instrument is badly needed.”
Addressing a hypothetical concern that a futures market might simply encourage reckless borrowing and overbuilding—allowing operators to borrow against depreciating hardware by using artificial paper hedges—Wilson firmly pushes back. A futures contract cannot clear in a vacuum; it requires two rational counterparties. Hedging demand from suppliers looking to lock in future revenue will naturally pull forward contract prices downward. To clear the market, prices must adjust to levels attractive enough to entice speculative buyers seeking return:
> “For a futures contract to exist, there has to be a willing buyer and a willing seller.”
- Anthropic's breakthrough release drove an immediate run on inference compute capacity. - Hardware prices (B200, B300, H100) surged concurrently with a collapse in token index pricing, demonstrating divergent pricing mechanics across compute inputs versus outputs. - Forward curve pricing will reflect natural structural depreciation and heavy selling pressure from capacity owners hedging capital expenditures.
---
`15:00` Market Heterogeneity, Token Futures, and the "Crunch Spread"
The discussion turns to whether the AI compute ecosystem is broad enough to support liquid exchange-traded futures rather than remaining reliant on private, bilateral off-take agreements between a handful of hyperscalers (Meta, Alphabet, Amazon) and frontier labs (OpenAI, Anthropic). Wilson argues that the explosive growth of open-source models has already created a sufficiently fat tail. Multibillion-dollar platforms like Fireworks AI and Together AI exist precisely to host open-source models on compute clusters and sell the output, proving that the ecosystem is ready.
When asked how a standardized futures market can accommodate disparate architectures—ranging from custom ASICs and alternative chips like AMD, Cerebras, Etched, AWS Trainium, and Google TPUs to differing software stacks and cluster sizes—Wilson explains that baseline liquidity in pure Nvidia H100s or B200s is already vast enough to anchor trading. Other hardware profiles and specialized setups (such as massive networked clusters for training runs) will trade as a basis spread against that liquid benchmark.
Looking forward, Wilson reveals DRW's conceptual work on token futures—derivatives tracking the actual output tokens produced by models. Tying compute inputs to token outputs mirrors traditional industrial processing economics:
> “So in oil, we call it a crack spread. In soybeans, we call it a crush spread. So you heard it here first. At DRW, we call this the crunch spread.”
```mermaid flowchart LR accTitle: The Compute Crunch Spread accDescr: Flowchart showing the economic transformation of raw GPU rental capacity into marketable tokens, creating the crunch spread. subgraph Inputs["Input: Compute Capacity"] A["Rent GPU Cluster<br>(H100 / B200 On-Demand)"] end subgraph Processing["Processing: Model Deployment"] B["Deploy Open-Source Model<br>(Fireworks / Together AI)"] end subgraph Outputs["Output: Token Generation"] C["Sell Generated Tokens<br>(OpenRouter / Silicon Data Index)"] end A --> B B --> C C -.->|"Processing Margin<br>(The 'Crunch Spread')"| A ```
- **Open-source infrastructure scale**: Platforms such as Fireworks AI and Together AI demonstrate deep demand for open-market inference outside the closed frontier labs. - **Basis trading**: Disparate accelerators (AMD, Cerebras, Etched, Trainium, TPUs) and high-tier training clusters will be priced relative to standard H100/B200 benchmark contracts. - **The "Crunch Spread"**: DRW's term for the processing margin between the rental cost of raw GPU compute and the market sales price of model-generated tokens, analogous to an oil crack spread or soybean crush spread. - **Data benchmarking**: Pricing discovery for output contracts is already being informed by emerging metrics like the Silicon Data Token Index.
`20:00` Structuring GPU Futures and Assessing NVIDIA's Dominance
The discussion turns back to the practical mechanics of launching GPU futures. The hosts ask whether every newly released microchip requires its own distinct contract and dedicated liquidity pool. Don Wilson confirms this view, acknowledging the apparent friction of keeping up with an aggressive technological cycle. However, he contextualizes the pace: while fast, “it's rapid, but not that rapid.” The market will not demand a futures contract for every single hardware SKU, but rather standard contracts concentrated on the major, dominant chips. To illustrate that a proliferation of contracts is normal for commodity markets, Wilson points to power trading: “Think about how many electricity contracts there are, for every little node and, I mean, you know, thousands of them.” The hosts agree, noting that while markets trade thousands of regional oil contracts, broad public discourse typically tracks only two primary benchmarks.
```mermaid flowchart TD accTitle: Chip Cycle to Commodity Futures Framework accDescr: Diagram showing how continuous chip cycles lead to targeted commodity futures contracts and broader market growth.
subgraph TechCycle["Hardware Landscape"] A["Rapid Chip Releases"] --> B["Identify Dominant Hardware Architectures"] end
subgraph Derivatives["Market Contract Design"] B --> C["Major Benchmark GPU Futures"] C --> D["Targeted Liquidity Pools"] E["Analogous Models: Power Nodes & Crude Grades"] -.-> C end
subgraph Industry["Hardware Market Dynamics"] F["NVIDIA Expansion"] --> G["Total Addressable Market Grows"] H["Rival Chipmakers Enter"] --> I["NVIDIA Market Share Normalizes"] end ```
The conversation pivots to whether NVIDIA’s near-monopoly on high-performance compute is vulnerable to competitors. When asked if rival chipmakers can realistically erode that dominance, Wilson separates gross company growth from percentage market share. As he evaluates the broader competitive landscape: “I think that the space is going to continue to grow, and I can't—it's unlikely that NVIDIA will be able to maintain the market share that they've maintained, but I'm pretty confident that NVIDIA will continue to grow as a company.”
- GPU futures contracts will focus on dominant chips rather than every hardware variant that enters the market. - Electricity markets provide a structural precedent, successfully supporting thousands of localized nodal contracts. - Wilson expects NVIDIA’s company valuation and total revenues to climb alongside an expanding compute pie, even as alternative chipmakers gradually chip away at its dominant percentage share.
---
`25:00` Internal AI Tooling and Token Spend Routing at DRW
Reflecting on their conversation from the previous year, the hosts revisit how artificial intelligence is deployed inside DRW. At the time, trading teams used AI agents to debate market hypotheses and evaluate risk allocation, but had stopped short of letting autonomous models directly execute or deploy trading risk. Asked what has changed in the intervening twelve months, Wilson notes that adoption has surged across the firm: “There's definitely more frontline adoption, for sure.”
This internal usage has driven an explosive climb in computational resources. When asked about the firm's overall budget for running models, Wilson notes that “the token use is like this exponential chart.” With employees across diverse trading and research desks naturally demanding top-tier access, the firm had to tackle the operational headache of internal token management. Rather than assigning unrestricted access to bleeding-edge tools, DRW focuses on model routing efficiency. Wilson explains the operational strategy: “Well, here, I mean I think the most important thing is to be thoughtful about how you route the inference requests. And not every inference request needs to go to a frontier model. And so, you know, that's the thing that we're trying to optimize for is minimize token spend by optimally routing inference requests.”
- Internal trading teams have significantly increased day-to-day, frontline utilization of AI workflows over the past year. - Aggregate enterprise token consumption continues to chart an exponential growth trajectory. - Cost and performance optimization relies on dynamic inference routing, ensuring low-complexity tasks bypass costly frontier models.
---
`26:15` The Case for Perpetual Futures: Roll Costs, Hard-to-Borrow Stocks, and Regulatory Hurdles
The dialogue shifts toward perpetual futures ("perps"), with the hosts noting that CME Chairman Terry Duffy is sitting in the live audience and has expressed skepticism about them. Acknowledging that he is one of Duffy's clients and enjoys debating the topic with him, Wilson mounts an economic defense of perpetual derivatives: “My view is that perps serve a useful economic purpose. They are futures contracts.”
To highlight their utility, Wilson points to retail-accessible exchange-traded funds (ETFs) that hold commodities like crude oil. Historically, oil ETFs rolled front-month contracts into subsequent expiries on rigidly predictable schedules, creating noticeable slippage and structural front-running before they shifted to rolling allocations across a three-month window. A perpetual contract tracking the benchmark would eliminate the need to constantly roll expiries, resulting in a cleaner and lower-cost product for retail ETF investors. While acknowledging that producers still require traditional dated contracts to hedge long-term output, Wilson argues that dated and perpetual contracts should exist simultaneously: “You can have monthly and quarterly futures and perps side by side. They actually augment one another and that's a good thing.”
Wilson presents a second compelling use case involving hard-to-borrow equities. Retail investors who hold shares targeted by aggressive hedge-fund short sellers typically see zero benefit from the exorbitant borrow fees (often around 100% annualized) collected by brokerage firms: “If you have that stock in your retail account, your broker is probably not saying to you, 'Hey, by the way, we're going to give you 300 basis points a day because you're holding this hard-to-borrow stock.' Okay, that would be 1,000% annualized, but 30 basis points a day. So you're not getting the benefit of owning that hard-to-borrow stock.” By holding a perpetual contract instead, the high borrow fee translates directly and mechanically into regular funding rate payments delivered straight to the long holder.
The hosts agree that while the word "democratization" is often an overused marketing buzzword, this dynamic truly opens up suppressed yields to retail participants. However, the discussion immediately bumps into the critical structural hurdle: US derivatives regulations have traditionally mandated that any instrument designated as a futures contract must have a fixed expiration date.
- Oil ETFs could avoid roll slippage and mechanistic inefficiencies by tracking perpetual instruments rather than continually rolling front-month futures. - Dated quarterly futures and undated perpetuals can coexist in the same asset class to serve both commercial hedgers and index trackers. - Retail equity holders in hard-to-borrow shares rarely share in borrow fees, whereas holding perpetuals passes funding payments directly through to the long position. - Regulatory definitions requiring fixed expiry dates remain the principal obstacle to launching US-compliant perpetual futures.
`30:00` Futures vs. Swaps and the Perpetual Contract Debate
A sharp disagreement emerges over the regulatory classification and risk profile of perpetual contracts, picking up on a debate sparked by CME Group's Terry Duffy. While Duffy characterized perpetuals as swaps, Don Wilson firmly contends they are futures. When the hosts ask why the difference between a swap and a future cannot be easily settled, Wilson recounts asking that exact question to congressional staffers drafting the Dodd-Frank Act for Barney Frank. The staffer offered the circular justification that everybody simply knows swaps are risky. Wilson rejects this regulatory distinction, arguing that clearing eliminates any meaningful structural difference: “A swap that is cleared has the same economic risk as to society to the holder as a futures contract does.” In Wilson's view, market participants and policymakers have completely lost their minds over mere nomenclature.
The debate turns to leverage and risk mechanics, specifically the criticism that perpetual contracts introduce dangerous instability into markets. Wilson pushes back against the pervasive assumption that perpetuals inherently require extreme leverage or predatory auto-deleveraging mechanisms. He stresses that these parameters are contract specifications rather than inherent traits of the perpetual format itself, pointing out that an exchange could just as easily design a standard quarterly futures contract with dangerous auto-deleveraging and massive leverage. Reframing the utility of the structure, Wilson argues: “You could list a perp with reasonable leverage and no auto-deleveraging, and it would be a very useful economic instrument.” Joe Weisenthal jokes that next year the hosts should simply remove themselves from the stage entirely and let Terry Duffy and Don Wilson debate regulatory classifications directly, marveling that the definition of a swap has become one of the podcast's most animated debates.
```mermaid flowchart TD accTitle: Cleared Derivatives Risk vs Mechanics Architecture accDescr: Diagram illustrating Don Wilson's argument comparing cleared swaps, perpetual contracts, and traditional quarterly futures.
subgraph LegalNomenclature ["Superficial Legal Categorization"] direction LR S["Cleared Swaps"] F["Standard Futures"] end
subgraph CoreRisk ["Identical Economic Reality"] CR["Equal Risk Profile<br/>Cleared through central clearinghouse"] end
subgraph ContractDesign ["Contract Design Choices"] P["Perpetual Futures (Perps)<br/>• Funding rate replaces expiration<br/>• Can operate with standard leverage<br/>• Does not require auto-deleveraging"] Q["Quarterly Futures<br/>• Fixed expiration date<br/>• Can also be structured with extreme leverage<br/>• Can also use auto-deleveraging"] end
LegalNomenclature --> CoreRisk CoreRisk --> ContractDesign ```
- **Regulatory History**: During the drafting of Dodd-Frank, congressional drafters defended treating swaps differently from futures simply by asserting that swaps were widely understood to be riskier, ignoring that central clearing standardizes systemic counterparty risk. - **Contract Design vs. Format**: Wilson clarifies that excessive leverage and automatic liquidation/deleveraging engines are exchange-level design choices commonly found in offshore crypto exchanges, not structural prerequisites of perpetual funding-rate contracts.
---
`31:30` The Inevitability of 24/7 Trading
The hosts ask Wilson about the cultural and operational toll of continuous, round-the-clock market trading as traditional finance edges closer to operating on weekends and late nights. Wilson immediately counters the framing that 24/7 trading is an upcoming shift, emphasizing that for firms like DRW, it has been reality for over a decade. Having actively traded cryptocurrencies since 2014, Wilson notes that his firm is already accustomed to risk moving at all hours of the weekend.
While market makers who specialize in continuous markets are adapted to the lifestyle, Wilson acknowledges that the transition creates genuine friction for traditional market makers who are unaccustomed to round-the-clock exposure. In particular, interest rate options market makers dread the thought of having to actively quote and manage SOFR options throughout Saturday and Sunday. Even so, Wilson views this transition as an unyielding structural evolution in market structure that participants will be forced to adapt to: “And you know what? I'm very sympathetic to that. You know what? It's just gonna happen. We're going to have to figure out how to deal with it.”
- **Trading History**: DRW has maintained active 24/7 trading desks and continuous risk operations since entering the cryptocurrency markets in 2014. - **Options Complexity**: SOFR options market makers face steep operational challenges in managing nonlinear volatility risk during off-market hours without the liquidity of primary underlying cash markets.
---
`32:45` Show Outro and Production Credits
Joe Weisenthal and Tracy Alloway wrap up the live episode recorded at City Winery in Chicago, thanking Don Wilson for returning to the stage and promising to make his appearance an annual tradition. The hosts direct listeners to follow the podcast team on social media and access Bloomberg's daily newsletter and community channels.
- **Hosts**: Tracy Alloway and Joe Weisenthal. - **Venue**: City Winery in Chicago. - **Production Team**: Carmen Rodriguez, Dashiell Bennett, Kael Brooks, and Kevin Lozano. - **Community Platforms**: The *Odd Lots* Discord community (`discord.gg/OddLots`) and the daily newsletter at `Bloomberg.com/OddLots`.
## Notable quotes > `26:15` My view is that perps serve a useful economic purpose. They are futures contracts. — Don Wilson
> `30:00` You could list a perp with reasonable leverage and no auto-deleveraging, and it would be a very useful economic instrument. — Don Wilson
> `26:15` You can have monthly and quarterly futures and perps side by side. They actually augment one another and that's a good thing. — Don Wilson
## Mentioned Don Wilson, Joe Weisenthal, Tracy Allaway, DRW, City Winery, Bloomberg, Odd Lots, CoreWeave, Anthropic, Compute Exchange, Silicon Data, CME Group, Terry Duffy, CFTC, IOSCO, White House, Nvidia, Nvidia H100, Nvidia B200, Nvidia B300, Meta, Alphabet, Amazon, OpenAI, Fireworks AI, Together AI, AMD, Cerebras, Etched, AWS Trainium, Google TPU, OpenRouter, Barney Frank, Dodd-Frank Act
## Video description <p>When we spoke to Don Wilson, founder and CEO of DRW, in Chicago last year, he told us he wanted to build a market for GPUs that would eventually be bigger than oil. A year's time in the age of AI is an eternity, and so for the first episode of this year's Chicago live show - recorded at City Winery - Wilson joins us on stage again to talk about what's changed and what hasn't in his quest to help build this market. We get his take on the recent tech selloff, including credit concerns around neocloud CoreWeave. And we also discuss how DRW is developing a “crunch spread” (the GPU version of a crack spread), the latest on his plans to launch GPU futures, and the possibility of token futures too.<br><br>Read more: <a href="https://www.bloomberg.com/news/articles/2026-10-16/jump-drw-bet-on-new-compute-market-as-ai-boom-faces-scrutiny?utm_medium=referral&utm_source=podcast&utm_campaign=odd_lots&utm_content=article">Jump, DRW Bet on New Compute Market as AI Boom Faces Scrutiny</a><br><br><strong>Only Bloomberg.com</strong> <strong>subscribers can get the </strong><strong><em>Odd Lots</em></strong><strong> newsletter in their inbox, plus unlimited access to the site and app.</strong> <a href="https://www.bloomberg.com/subscriptions/oddlots?in_source=oddlotspodcast"><strong>bloomberg.com/subscriptions/oddlots</strong></a></p><p>See <a href="https://omnystudio.com/listener">omnystudio.com/listener</a> for privacy information.</p>
[Source](https://podtrac.com/pts/redirect.mp3/tracking.swap.fm/track/UVBrz8bN8aM2Xe47PEPu/traffic.omny.fm/d/clips/e73c998e-6e60-432f-8610-ae210140c5b1/8a94442e-5a74-4fa2-8b8d-ae27003a8d6b/80dab570-fb6b-445e-b20e-b4e5015a3603/audio.mp3?in_playlist=982f5071-765c-403d-969d-ae27003a8d83)