⛽ Trading compute
1. Key Themes
Theme 1: Compute Is Becoming a Tradeable Commodity Asset Class
Wall Street is moving to financialize AI compute infrastructure, treating GPU capacity the way commodity traders treat oil or jet fuel — with futures contracts, price benchmarks, and hedging instruments.
"Investors increasingly want to trade compute like a commodity, betting it could make the historically expensive AI buildout more sustainable and efficient."
"Goldman Sachs estimates that between 2026 and 2031 roughly $7.6 trillion will be invested globally in building up compute, power and data centers. But the financial infrastructure needed to sustain that level of spend 'has not yet been built.'"
Theme 2: Government Is Now a De Facto Gatekeeper for Frontier AI Model Releases
The Anthropic episode reveals a new reality: the U.S. government can effectively block or delay release of the most powerful AI models, and AI labs must proactively manage that relationship.
"Anthropic eventually understood that in order to be successful, it needed to be on the same side as the government."
"There's a lot of work left to be done on a framework for approving future models with a clear inclusive process that has transparency standards and timelines."
Theme 3: AI Model Safety Standards Are Undefined and Inconsistently Applied
The jailbreaking concern that triggered export controls was disputed by independent experts, exposing that there is no agreed-upon safety benchmark for what constitutes a releasable model.
"Cybersecurity experts, however, later wrote in an open letter to the administration that other leading AI models have the same issue Amazon warned about with Anthropic."
"The federal Center for AI Standards and Innovation and the National Security Agency said those changes weren't good enough, prompting further fixes."
Theme 4: The AI-China Competition Is Shaping Domestic Regulation and Deployment
Both the compute commoditization story and the Anthropic model saga are framed through the lens of U.S.-China competition, making geopolitics a first-order constraint on AI business strategy.
"His co-founder and CEO, Kush Bavaria, sees this as part of America's potential advantage over China and added that the company does not work with Chinese AI labs."
"It remains uncertain when and how Anthropic's models will be released to ally countries around the world — which proponents say is key to beating China."
2. Contrarian Perspectives
Compute Cannot Be Commoditized Like Oil — The Asset Is Fundamentally Perishable and Depreciating
The article makes clear that while Wall Street wants to trade compute like a commodity, the underlying asset has structural properties that make standardized futures contracts extremely difficult. This is a significant risk for investors betting on this infrastructure layer.
"Each new generation of Nvidia chips promises better price performance, which changes the value of older chips. Any benchmark has to chase a depreciating asset."
"GPU capacity can't be stored, so unused compute disappears, making it harder to build standardized contracts and pricing around it."
Amazon's Safety Alarm — The Trigger for a 20-Day Crisis — Was Disputed by Experts
The conventional read of the Anthropic shutdown is that it was a justified national security response. The contrarian read: a major investor (Amazon) may have set off a disproportionate regulatory action based on a security concern that was not uniquely attributable to Anthropic's models, raising questions about competitive motivations and the reliability of corporate-to-government security disclosures.
"It warned about a 'jailbreaking' issue it found with the AI lab's latest models... Cybersecurity experts, however, later wrote in an open letter to the administration that other leading AI models have the same issue Amazon warned about with Anthropic."
The Real AI Policy Power Broker Is the Treasury Secretary, Not Tech-Focused Agencies
The instinct is to watch Commerce, NSA, or OSTP for AI policy signals. The article suggests Treasury Secretary Scott Bessent was the pivotal figure — the first call Andy Jassy made, and the official who brokered the resolution.
"Out of all of the administration officials Amazon's Andy Jassy could have called, it was Treasury Secretary Scott Bessent who first heard about the jailbreaking issue."
"Bessent was early to sound the alarm on Mythos, work with White House chief of staff Susie Wiles to reengage the embattled company and help get a cybersecurity executive order across the finish line."
3. Companies Identified
- Description: Andreessen Horowitz-backed startup building a marketplace for trading GPU compute
- Why mentioned: Lead company in the compute commoditization story; raised a $33M seed round
- Quote: "Ornn, an Andreessen Horowitz-backed startup, raised a $33 million seed round to build a marketplace for trading the computing power that underpins today's AI boom, similar to what exists for oil traders."
Anthropic
- Description: AI safety-focused lab, maker of models Mythos and Fable
- Why mentioned: Central to the model shutdown/revival story; spent 20 days offline under Trump administration export controls
- Quote: "Anthropic's models are back online, but the impact of its 20-day showdown with the Trump administration will be long lasting."
Amazon
- Description: Cloud giant; investor and partner in Anthropic
- Why mentioned: Triggered the government intervention by flagging the jailbreaking issue to Treasury Secretary Bessent
- Quote: "It warned about a 'jailbreaking' issue it found with the AI lab's latest models, Mythos and Fable... Amazon flagged its concerns to the administration, triggering sweeping export controls."
- Description: Global investment bank
- Why mentioned: Provided the $7.6 trillion compute investment estimate and noted the absence of supporting financial infrastructure
- Quote: "Goldman Sachs estimates that between 2026 and 2031 roughly $7.6 trillion will be invested globally in building up compute, power and data centers."
CME (Chicago Mercantile Exchange)
- Description: Major global derivatives exchange
- Why mentioned: Plans to launch compute futures tied to Silicon Data's benchmark, pending regulatory approval
- Quote: "Pending regulatory approval, CME plans to launch compute futures tied to Silicon Data's benchmark."
Intercontinental Exchange (ICE)
- Description: Global exchange operator
- Why mentioned: Plans to launch GPU compute futures tied to Ornn's pricing index
- Quote: "The Intercontinental Exchange plans GPU compute futures tied to Ornn's pricing index."
- Description: Leading AI lab, maker of GPT series models
- Why mentioned: Its latest model GPT-5.6 is on hold, and it was excluded from the Anthropic-White House discussions
- Quote: "OpenAI, whose latest model GPT-5.6 is on hold, did not have visibility into discussions between Anthropic and the White House and is engaged in daily technical discussions on the release of its own model."
- Description: Chinese tech giant, parent of TikTok
- Why mentioned: Making inroads in Hollywood with Seedance, its AI video creation tool
- Quote: "ByteDance is making inroads in Hollywood with the quality and low cost of its Seedance AI video creation tool."
Amazon Mechanical Turk
- Description: Amazon's gig-work crowdsourcing platform
- Why mentioned: Shutting down to new customers, signaling the end of human-powered data labeling
- Quote: "Amazon said it is no longer accepting new customers for Mechanical Turk and is placing the gig work system in maintenance mode, a signal it could shutter it entirely."
4. People Identified
- Description: CTO and co-founder of Ornn
- Why mentioned: Articulated the company's vision for AI compute financing
- Quote: "We want to make financing AI way more seamless."
Kush Bavaria
- Description: CEO and co-founder of Ornn
- Why mentioned: Framed compute trading as a U.S. competitive advantage over China
- Quote: "His co-founder and CEO, Kush Bavaria, sees this as part of America's potential advantage over China and added that the company does not work with Chinese AI labs."
Dario Amodei
- Description: CEO of Anthropic
- Why mentioned: Directly received the export control letter from Commerce Secretary Lutnick; led Anthropic through the crisis
- Quote: "Amodei called Lutnick back that night after receiving the letter, realizing it effectively meant the models would have to be taken offline."
Howard Lutnick
- Description: U.S. Commerce Secretary
- Why mentioned: Directed the export control action against Anthropic and was at the center of the technical resolution process
- Quote: "Lutnick made clear to Amodei the issue needed to be resolved fast and alerted the CEO that the company would be receiving a letter imposing sweeping export controls."
Scott Bessent
- Description: U.S. Treasury Secretary
- Why mentioned: Was the first government official to hear about the jailbreaking issue; played a key brokering role in the resolution
- Quote: "Bessent was early to sound the alarm on Mythos, work with White House chief of staff Susie Wiles to reengage the embattled company and help get a cybersecurity executive order across the finish line."
Tom Brown
- Description: Co-founder of Anthropic
- Why mentioned: His technical depth allowed him to negotiate directly with government security specialists
- Quote: "Brown's technical expertise allowed him to sit in a room with government specialists and go line by line through how models behave under stress."
Sarah Heck
- Description: Anthropic's policy chief
- Why mentioned: Became more involved as discussions turned more technical during the government standoff
- Quote: "As discussions turned more technical, Anthropic policy chief Sarah Heck and Anthropic co-founder Tom Brown got more involved."
Andy Jassy
- Description: CEO of Amazon
- Why mentioned: Chose to call Treasury Secretary Bessent — not a tech or national security official — as his first government contact on the issue
- Quote: "Out of all of the administration officials Amazon's Andy Jassy could have called, it was Treasury Secretary Scott Bessent who first heard about the jailbreaking issue."
Sean Cairncross
- Description: National cyber director
- Why mentioned: Participated in technical discussions during the Anthropic resolution process; had direct conversations with Anthropic's Tom Brown
- Quote: "National cyber director Sean Cairncross... also participated in technical discussions."
Sam Corcos
- Description: Treasury Department chief information officer
- Why mentioned: Participated in the multi-agency technical discussions resolving the Anthropic standoff
- Quote: "Treasury Department chief information officer Sam Corcos and the NSA also all participated in technical discussions."
Susie Wiles
- Description: White House chief of staff
- Why mentioned: Worked with Bessent to reengage Anthropic during the crisis
- Quote: "Bessent was early to sound the alarm on Mythos, work with White House chief of staff Susie Wiles to reengage the embattled company."
5. Operating Insights
AI Companies Need a Government Relations Strategy That Goes Beyond Policy Teams
The Anthropic saga shows that technical co-founders — not just policy staff — need to be able to engage directly with government security specialists. Anthropic's breakthrough came when Tom Brown entered the room.
"Brown's technical expertise allowed him to sit in a room with government specialists and go line by line through how models behave under stress."
Takeaway: For any AI company building frontier models, the ability to speak the government's technical language — not just lobby on policy — is now a core operational competency.
Compute Procurement Strategy Is a Financial Risk Management Problem, Not Just an Engineering One
AI companies currently manage compute exposure through long-term pre-purchasing agreements, a blunt instrument. The emerging futures market suggests CFOs and finance teams should be thinking about compute hedging the way airlines think about jet fuel.
"AI companies don't have an equivalent market for compute, a problem the 20-something founders of Ornn and a growing number of exchanges are trying to solve. So far, AI companies have tried to lock up supply and prices through long-term pre-purchasing agreements."
Takeaway: Operators building compute-intensive products should watch CME and ICE futures launches closely — hedging tools could materially change how AI infrastructure costs are managed and budgeted.
6. Overlooked Insights
Ornn Is Already Integrated With Bloomberg Terminal — Giving It Embedded Distribution Before Its Market Even Launches
Most early-stage marketplace startups struggle with distribution. Ornn has quietly solved this by plugging into the terminal that every institutional trader already uses daily — a significant competitive moat before regulatory approval is even granted.
"Ornn has already integrated with Bloomberg Terminal and other data providers, which lets traders check GPU prices through the tools they already use."
The Model Freeze Is Not Limited to Anthropic — OpenAI's GPT-5.6 Is Also on Hold
The article treats the Anthropic story as the headline, but buries the fact that OpenAI's latest model is also currently frozen pending its own government review. This suggests a systemic new regulatory regime affecting all frontier labs, not a one-off incident.
"OpenAI, whose latest model GPT-5.6 is on hold, did not have visibility into discussions between Anthropic and the White House and is engaged in daily technical discussions on the release of its own model."