Posts

On 28 September Reuters reported that two customers accounted for nearly a quarter of Anthropic's revenue in 2025. The reporting that followed points at something larger. Anthropic's filing is still confidential, and a search of SEC EDGAR on 5 October found no public registration statement, but Reuters, which has seen it, reports the share of revenue sold through Amazon and Google rising from 11% in 2023 to 32% in 2024 and 47% in 2025, about $2.16 billion, through their cloud marketplaces. The reading here is that the larger concentration is the channel rather than the customer list. Nearly half of last year's revenue arrived through two marketplaces, and the companies that own them are, by Reuters' account of the filing, also Anthropic's investors, its compute suppliers, its competitors and its customers. Reuters still does not name the two customers who each generated 12% of revenue in 2025, and this piece does not guess. If either of them is one of those same two companies, the two concentrations fall on the same names and add up to more than either figure shows on its own.

On 12 September the chief executive of Anthropic published an essay arguing that the industry should slow the rate at which it improves model capabilities. The heads of OpenAI, xAI and Google DeepMind all responded in public within a day, and the coverage has treated that as three rivals agreeing. Agreement on the direction is not agreement on the mechanism, and on the mechanism they do not agree. This piece does not judge whether the argument is right. It examines the first of the three proposed steps, because that step is a disclosure mechanism, and disclosure mechanisms are readable from here. Each frontier company would give ongoing, employee-like access to a team of embedded third-party evaluators, covering not just finished models but training pipelines and processes. Strip the subject matter away and the shape is familiar. This is a proposal to install an independent party inside the building, with standing access and the right to publish what it finds.

The AI build is usually discussed as a technology story. But underneath it sits a capital-allocation question that may matter just as much: who is actually structured to hold the risks being created? This first piece in a 3-part series starts at the bottom of the stack with basic scientific research. It is long-duration, failure-prone, and difficult for the company funding it to fully monetize. Historically, that has made it an awkward fit for private balance sheets. Yet frontier AI companies are spending heavily on exactly this kind of work. The article examines why basic research behaves like the first-loss equity tranche of technological development, why neither governments nor private companies are obvious natural holders of that risk, and what the current AI investment cycle may be telling us about a decades-old economic assumption.

SpaceX reported $18.37 billion in quarterly capex. Two days later, it announced the first phase of a semiconductor campus that could eventually involve far more spending. The difference between those numbers is more important than the market reaction. Committed spending gets measured. Announced spending keeps its optionality. That matters as AI infrastructure spending accelerates toward extraordinary levels and a growing share is financed with debt. It matters even more as frontier AI companies prepare to enter public markets. The deeper question is whether quarterly reporting is built to price research programs whose costs arrive now but whose payoff may be decades away. SpaceX offers a useful case study in what happens when long-duration technological bets meet the short-duration accountability of public markets.

The market interpreted the reported OpenAI delay primarily through a macro lens, suggesting that the mega-IPO wave was slowing due to weakening public demand. However, this perspective may not be the most insightful. A more effective lens to consider is structure. If the reports are accurate, the inquiry posed to OpenAI was not about whether to go public at all, but rather whether to list sooner at a valuation below $1 trillion or to wait for market conditions that would support its desired valuation. Sam Altman reportedly viewed a reduction to that figure as unacceptable. This distinction is significant because it separates two concepts that are often conflated: investor appetite and entry point. A headline stating "OpenAI delays IPO" might imply a decrease in demand, but it could actually indicate a commitment to price discipline. The public market may still be interested in the company, but it may not be willing to meet the private-market valuation being proposed, which conveys a very different message.

Two weeks ago, the market's loudest fear was supply. SpaceX had just executed the largest IPO in history, raising approximately $75 billion at $135 a share, valuing the company near $1.77 trillion. Following closely were OpenAI and Anthropic, both anticipated to follow suit. The bearish narrative spread quickly: a wave of mega-listings could extract a trillion dollars of value from the broader market as investors sold existing holdings to fund new shares. However, this week, the anticipated wave began to slow down. OpenAI is reportedly considering delaying its IPO until 2027. Advisers have presented the company with a choice: list sooner at a valuation below $1 trillion or wait for more favorable conditions that align with its desired valuation. Sam Altman reportedly shows little interest in accepting a discount. The news caused a significant drop for SoftBank, one of OpenAI’s largest outside backers. This caution signals a shift in focus. The key question is no longer whether demand exists for these listings, but whether that demand will support the valuations seen in the private market.

T-OpenAI represents a tokenized loan participation right, offering economic exposure tied to the valuation of OpenAI private shares. It is important to note that T-OpenAI is not equity; it does not grant the holder shareholder status, ownership, voting rights, or dividends. The overarching goal is straightforward: to move the access point to this type of investment earlier. This financial instrument is structured through a Cayman Islands SPC and issued by a dedicated subsidiary of Tessera Works Foundation. It operates entirely on the Solana blockchain, ensuring liquidity, transferability, and composability from day one. The unique structure of T-OpenAI enables private-market exposure, which has traditionally been limited to institutional investors, to be brought on-chain. This approach provides clarity regarding the nature of the asset and its limitations.

OpenAI’s latest private funding round in March 2026 has set the company's valuation at approximately $852 billion. This valuation gap is significant, as it highlights the difference between OpenAI's current private mark and the anticipated IPO target, which is already in the hundreds of billions. However, a major challenge remains: access to OpenAI shares is limited. Currently, they are traded in a closed private market, primarily available to insiders, large funds, and accredited investors with the necessary connections. For the majority, the opportunity to engage with OpenAI may mirror the experience many had with SpaceX, arriving only after the market has already adjusted its valuation. Tessera changes this.

The line of mega-IPOs forming behind SpaceX includes a key player we have been closely monitoring, OpenAI. The significance of OpenAI is clear: it continues to hold the front door to AI technology. For many, ChatGPT represents their first substantial interaction with AI, which carries immense importance. The product reportedly engages around 900 million weekly users, maintaining a lead over any other AI-native product in terms of direct consumer reach. Although growth has slowed compared to the company's internal expectations, this should be taken seriously. Nevertheless, no competitor has matched the unique combination of scale, familiarity, and direct user engagement that OpenAI established first. This extensive reach does more than attract attention; it fosters distribution and builds a brand that the public recognizes. OpenAI also enjoys the broadest revenue opportunities among frontier labs, including consumer subscriptions, enterprise seats, and an API layer that supports thousands of downstream products. While many companies are developing impressive models, far fewer maintain a global relationship with end users at such scale.

Behind it are the AI labs. Two of the leading frontier AI companies are reportedly preparing 2026 listings and positioning themselves behind SpaceX. One appears to have moved first with a confidential filing. The other seems willing to wait, likely giving itself room to price against whatever the earlier deals establish. Together, they could be seeking well over $200 billion in proceeds from the same broad pool of capital, inside the same general window. That is where the real constraint starts to show. IPO timing is never chosen in isolation. It is chosen relative to every other company asking the market for money at the same time.