2026-05-22-45-mins-dustin-ross-and-spencer-powers

Debrief: 45 Mins Dustin Ross And Spencer Powers — 2026-05-22

Summary

Dustin met with Spencer Powers, a DRW compute portfolio builder who co-founded companies before creating a hybrid incubation/investment/trading role at DRW. The conversation centered on the financialization of compute markets — specifically the thesis that AI and tech companies lack the hedging infrastructure that commodity industries like oil have long relied on. Spencer shared candid views on why tech CFOs will be slow to adopt futures products and where the real near-term demand will come from, and closed by offering an introduction to Joseph at Compute Exchange.

Key Themes

1. Financialization of Compute Markets: The Oil Analogy and Its Limits Spencer validated Dustin’s core thesis directly — crediting Don Wilson with articulating the same framing in September 2023: ‘the financial and risk infrastructure that oil has that compute doesn’t.’ DRW has been operationalizing this thesis for two years, and the Silicon Data / CME futures launch two weeks prior represents the first real proof of concept. However, Spencer immediately complicated the thesis by questioning demand: ‘That is our big risk on launching Futures, right? Will people actually use them?’ The conversation did not resolve this — it surfaced the adoption gap as the central open question. Spencer’s framing of dollar-per-GPU-hour as the commoditization unit is important: it encompasses power costs, avoids the GPU monopoly pricing problem (NVIDIA), and reflects the actual language spoken between neoclouds and their customers. He also flagged a potential future shift toward dollars-per-token as agentic frameworks mature, though he treated this as a segment-by-segment evolution rather than an immediate displacement.

2. Tech Industry’s Lack of Financial Sophistication: The COGS Blindspot This was the most energized thread in the conversation. Spencer’s observation that ‘no CFO of a tech company has ever had to hedge their cost of goods sold’ landed as a genuine insight for Dustin, who echoed it back. The underlying dynamic: SaaS historically had ~99% margins with near-zero infrastructure costs, so financial risk management around COGS simply wasn’t a discipline that developed. AI products break this model — inference costs are real, they scale with usage, and they’re subject to commodity price swings. Spencer added a second layer: AI executives are launching products without knowing demand, compounding the volatility. ‘It’s impossible for them to have scalable inference in the sense that price doesn’t change.’ This creates a structural education gap that any hedging product or advisory service would need to solve before it could sell.

3. Debt Financiers as the Beachhead Adopter Spencer made a clear prediction that was more actionable than the general thesis: providers of debt capital to neoclouds will adopt futures hedging before tech companies do. His reasoning is direct — if the assets collateralizing their loans are compute infrastructure, and compute prices are volatile, they have a clean one-to-one risk transfer motive. ‘I think at least providers of debt capital will be smart enough to hedge their exposure if the assets they’re collateralized against lose value.’ This is a more tractable buyer segment than AI company CFOs and maps to Wall Street / infrastructure finance rather than Silicon Valley.

4. Opportunity for Broker/Advisory Businesses Spencer raised this unprompted as a gap DRW’s portfolio companies won’t fill: ‘There probably is some opportunity for brokerages or advisors that package hedge products or work with the data centers and work with AI companies to hedge their cost of goods sold.’ He acknowledged it ‘might not be the sexiest startup’ and floated a consultative model, but framed it as structurally analogous to investment banking. Dustin noted this directly: ‘It basically sounds like an investment bank.’ Neither resolved whether this is a software product, a services firm, or a hybrid — it remained an identified opportunity without a clear structure.

5. Commodity Risk Anecdote: Storage Neocloud CEO Spencer grounded the abstract thesis with a concrete failure case: a neocloud CEO whose storage business was destroyed when memory costs rose 500%, forcing him to route customers to hyperscalers at a loss. ‘Right business plan destroyed by commodity price shifts.’ This is directly analogous to a farmer without crop insurance or an airline without fuel hedges — the tools didn’t exist, or the operator didn’t know to use them. This anecdote is the most transferable illustration of why hedging infrastructure matters, and it touches semiconductor supply chain risk (memory pricing) not just compute leasing.

6. DRW Portfolio Mechanics and Deal Flow Constraints Spencer described DRW’s four compute-adjacent bets — Silicon Data (futures data layer, CME partnership), Compute Exchange (spot market for reserved compute), Vast.ai (‘Airbnb for GPUs’), and SF Compute (cluster bursts for smaller startups). He flagged deal flow as DRW’s binding constraint: Chicago location and non-traditional VC identity create credibility headwinds with founders. He’s actively looking for high-quality pipeline, which is why the Stanford connection with Dustin has mutual value.

Notable Quotations

“No CFO of a tech company has ever had to hedge their cost of goods sold because their cost of goods sold was like nothing. Old school SaaS had 99% margins. Now AI products actually do have cost of goods sold in their inference cost.” — Spencer Powers. Context: Core insight on why tech companies will lag in adopting futures hedging — the discipline simply never developed.

“That is our big risk on launching Futures, right? Will people actually use them?” — Spencer Powers. Context: Spencer candidly naming DRW’s single biggest execution risk on the Silicon Data / CME launch — demand uncertainty, not supply.

“Right business plan destroyed by commodity price shifts.” — Spencer Powers. Context: Describing a storage neocloud CEO hit by 500% memory cost increases — the most concrete illustration of semiconductor supply chain risk in the conversation.

Themes & Contradictions

This conversation primarily confirms and extends prior threads rather than contradicting them. The oil/semiconductor financialization analogy — which Dustin and Bliss have been developing with Professor Steve and which appears as Thesis III in both the GEMINI and CLAUDE venture selection memos — received direct external validation from Spencer, who credits Don Wilson with the same framing independently in September 2023. This is meaningful triangulation: the thesis is not idiosyncratic.

However, this meeting surfaces a tension with the AI synthesis memos. Both GEMINI and CLAUDE rank the compliance wedge (Thesis I/IV) as the highest short-term opportunity, in part because the CLAUDE memo explicitly flags historical failures of semiconductor benchmarking — DRAM futures in the 1980s-90s and Enron’s 2001 forward contracts. Spencer did not address these failures, and the conversation implicitly treated the futures launch as progress rather than a repeat of prior attempts. Bliss and Dustin should hold that historical failure data against Spencer’s optimism.

The Richard Dasher meeting (November 2025) touched on critical materials dependency and US competitiveness but did not explore financialization directly. The Andras contact introduced the German Supply Chain Due Diligence Act as a regulatory tailwind — a compliance-driven entry point that sits in a different quadrant from Spencer’s market-structure argument. These two tracks (compliance mandate as wedge vs. financial infrastructure as wedge) are not yet reconciled in the corpus.

The internal session (P0004) flagged that no VP Export Compliance or trade counsel has been interviewed despite the compliance thesis ranking highest. This meeting does not close that gap — Spencer is a market-structure and trading infrastructure contact, not a compliance buyer. The interview coverage imbalance identified in P0004 remains unresolved after this conversation.

Business Problems & Painpoints

Spencer did not present himself as a buyer with direct pain — he is an operator building infrastructure. But he surfaced pain on behalf of market participants with unusual specificity:

Neocloud operators facing commodity price shock: The storage CEO anecdote is the most visceral pain point — a business plan executed correctly but destroyed by a 500% memory cost spike with no hedging mechanism available. This is not hypothetical; it already happened to a named operator. The pain is retrospective but the structural cause persists.

AI company CFOs managing inference cost volatility: Spencer described a compounding problem — new products launched without demand visibility, compute prices spiking when new models or agentic frameworks release, and no institutional tooling to manage it. ‘It’s impossible for them to have scalable inference in the sense that price doesn’t change.’ The pain is real but the buyer doesn’t yet know they have a solvable problem — the education gap is itself part of the friction.

Futures adoption drag: Spencer flagged that tech companies are ‘not financially sophisticated enough to understand the hedging benefits’ — which is simultaneously a pain point (unmanaged risk) and a go-to-market obstacle (unrecognized need). This is a harder sales motion than selling to buyers who already feel pain.

DRW’s own deal flow constraint: Spencer named this directly — Chicago location and non-traditional VC identity limit the quality of startup pipeline reaching them. This is Spencer’s personal workflow pain and the reason the Stanford relationship has value to him. Not relevant to the startup thesis but useful for the relationship.

The implied willingness to pay is highest among debt financiers (collateral protection motive is direct and quantifiable) and lowest among AI company CFOs (education required before value is recognized).

Emotional Signals

Spencer came across as confident and fluent — this is clearly his domain and he has been thinking about it for two years. He was energized when Dustin mirrored the oil/semiconductor framing back to him, and became most animated discussing the CFO COGS blindspot — that thread had a ‘yes, exactly’ quality from both sides. He was candid about DRW’s risks (‘will people actually use them?’) in a way that suggested comfort with uncertainty rather than evasion. Dustin’s Ukraine background visibly surprised Spencer and shifted the register of the conversation briefly — Spencer’s ‘be safe in Ukraine’ response was genuine. No frustration or defensiveness was apparent. Spencer’s most guarded moment was around the broker/advisory opportunity — he framed it as ‘not sexy’ twice, suggesting he sees the gap but doesn’t want to own it. The meeting felt like two people who agree on the macro and are trying to find where their specific interests overlap.

For Founders

1. Spencer named debt financiers to neoclouds as the most financially sophisticated near-term adopters of compute hedging — ahead of the AI companies and data centers you’ve been thinking about. Does that buyer segment connect to either of your two workstreams (supply chain intelligence vs. financialization), or does it point somewhere new?

2. Both AI synthesis memos flag the historical failure of semiconductor futures (DRAM in the 1980s-90s, Enron in 2001) as a reason to be cautious about the hedging thesis, but Spencer didn’t address those failures at all. How much weight should you put on the ‘it’s different this time’ argument given that DRW — arguably the most credible possible actor — is now betting on it?

3. The broker/advisory model Spencer described (packaging hedge products, educating COGS-blind tech CFOs) was explicitly identified as an opportunity DRW won’t pursue — but Spencer called it ‘not sexy’ twice and framed it as consultative. Does that gap look like a wedge into the financialization thesis, or is it a signal that the market isn’t ready for a software-first product in that space yet?