Raison Venture Digest

Anthropic is preparing for an IPO before Thanksgiving, Starship has reached orbit for the first time, and Lambda plans to raise up to $4 billion before going public. In this new issue, we cover the key events in the Raison portfolio and examine how control over AI agents is emerging as a separate market segment.

Anthropic

IPO before Thanksgiving, financial metrics, and control of the company

Anthropic expects to go public before Thanksgiving. In 2026, it falls on November 26. According to Bloomberg, the Claude developer will begin offering shares to investors around November 9. The expected valuation exceeds $2 trillion.

The company filed for an IPO confidentially, but Reuters and The Wall Street Journal have already reviewed a preliminary version of the S-1 prospectus. Here is what the document shows.

In the first half of 2026, Anthropic’s revenue reached approximately $16.3 billion, compared with $1.1 billion in the same period a year earlier. Along with revenue, spending on computing power and infrastructure also increased, from $2.3 billion to nearly $9 billion.

In the first half of the year, the company recorded a net loss of $80.7 billion. Of that amount, $72 billion came from an accounting revaluation of convertible loans and other financing instruments.

The main future expense item is computing power. Under contracts with cloud providers, the company will need to pay about $450 billion over 7–10 years, The Wall Street Journal reported. This amount could rise to $750 billion in the coming years.

Anthropic’s revenue depends heavily on a small number of large clients: about one-fifth of half-year revenue came from two of them. The company expects a significant share of revenue to continue coming from a limited group of clients.

At the same time, Anthropic is deciding who will control the company after the IPO. According to The Information, the company is asking shareholders to approve a share structure under which seven co-founders will receive 50.1% of the votes on most corporate matters. Each of them currently owns about 2% of the company.

Lambda

Plans for a new round and valuation growth to $14.5 billion

Lambda plans to raise up to $4 billion at a $14.5 billion pre-money valuation. This round could be the last private round before the company’s planned 2027 IPO. Earlier, Lambda discussed raising $3 billion at a valuation starting at $12 billion.

On October 1, Lambda also secured a loan of $1 billion at 6.78%, maturing in 2033. The funds will be used to purchase GPUs for three projects already under contract.

According to The Wall Street Journal, Lambda’s order book grew from $15 billion in June to $50 billion in September. Almost all of the increase came from one contract: at the end of August, the company signed an agreement with Anthropic worth $35 billion. Whether Lambda can justify the new valuation will largely depend on how it executes this contract.

The capital raised finances infrastructure built for decades, not quarters.

Lambda CEO

Michelle Combs

SpaceX

First orbital flight of Starship and expansion of NASA contract

On September 28, Starship, SpaceX’s largest rocket, reached Earth orbit for the first time. During its 14th test flight, Starship deployed 26 third-generation Starlink satellites into orbit. Each adds about 1 Tb/s of network capacity and improves service quality for subscribers.

SpaceX’s cooperation with NASA also expanded. In September, the agency increased the contract for crewed flights to the ISS by $946 million and added three new missions to it. The total contract value rose to $5.92 billion.

Crusoe

New data center for Google in Texas

On September 28, Crusoe confirmed that it began building a data center for Google in Texas in 2025. More than 5,500 people are currently working at the site every day.

The site is being built next to a 531 MW wind farm, which will power the campus directly, without waiting for connection to the state grid.

According to TheEnergyMag, Armstrong County, where construction is underway, is included in Google’s $40 billion investment plan in Texas, announced in November 2025.

To build the future of AI, we need a new kind of infrastructure, and it starts with energy.

CEO and co-founder of Crusoe

Chase Lochmiller

Cohere

Merger agreement with Aleph Alpha and talks on a new round

On September 16, Cohere and German language model developer Aleph Alpha signed a merger agreement, formalizing the April arrangements.

The combined company will operate under the Cohere brand, with headquarters in Toronto and Berlin. Aleph Alpha’s current headquarters in Heidelberg will become a research center. Closing is expected by year-end, subject to regulatory approval.

Schwarz Group, owner of Lidl and Kaufland, also supports the merger and plans to invest $600 million in the company. At the same time, according to The Globe and Mail, Cohere is in talks to raise another $2–3 billion, including from the Canadian government. We wrote about these talks in the previous issue.

The combined company’s valuation will be $20 billion, nearly three times higher than in September 2025. At that time, Cohere raised investment at a $7 billion valuation. By the end of 2025, its annual recurring revenue (ARR) reached $240 million.

Cohesity

New service for recovering AI agents after failures

Cohesity introduced Agent Resilience, a service for recovering enterprise AI agents after failures. Such agents work with companies’ internal systems and data, so their errors directly affect the business. Agent Resilience saves working versions of agents: if an error occurs, a company can quickly roll back to one of them and reduce losses.

The service is currently compatible only with Amazon Bedrock, the cloud platform for AI applications, and is available to a limited group of customers. Cohesity plans a full launch by year-end.

According to Cohesity’s own survey, 56% of organizations admitted they are not sufficiently prepared to detect and stop unwanted agent actions.

Detection systems will show that an AI agent is behaving incorrectly, but they will not be able to undo the changes it has made

Chief Product Officer, Cohesity

Vasu Murthy

Kraken

Talks with BNY and expansion of cooperation with Nasdaq

According to CoinDesk, Payward, Kraken’s parent company, is in talks with BNY on a financial infrastructure partnership. The parties are discussing custody, trading, payments, and wealth management. For Payward, this is a strategic move beyond being a crypto exchange: the company is investing billions of dollars in building infrastructure for the financial market.

At the same time, Payward is expanding its cooperation with Nasdaq. Earlier, the exchange invested $100 million in Payward at a valuation of $21 billion, according to Bloomberg. In the second quarter of 2027, the parties plan to launch tokenized Nasdaq Equity Tokens, and Payward will implement Nasdaq’s trading surveillance system on its platforms.

Neuralink

Progress in controlling a computer with the power of thought

Neuralink said that its neural implant, which allows people with paralysis to control a computer with their thoughts, set a record cursor-control speed of 11.32 bits per second. The record was set by the company itself.

In addition, the implant now needs to be adjusted less often. Neuralink trained the base model on more than 50,000 hours of brain activity recordings from clinical trial participants. As a result, some decoders that translate brain signals into commands can now operate for weeks without recalibration. The company published detailed results in a technical report.

Signal of the period

Control over AI agents is becoming a separate market

In the previous issue, we noted that AI safety is becoming an increasingly important factor for businesses and investors. Over the past month, this trend has developed further: new partnerships, products, and financial commitments have emerged.

On September 18, Anthropic announced that specialists from Accenture Faculty, Accenture’s AI division, will join its teams and test models and safety mechanisms. The partners plan to invest at least $1 billion in the project over five years. After the announcement, Accenture shares rose 8% in after-hours trading.

The need for such measures is also confirmed by incidents. Anthropic said it had identified four cases in which its models gained unauthorized access to third-party systems during testing. In addition, researchers participating in OpenAI’s bug bounty program used Claude to break into the company’s systems and gained access to ChatGPT and Codex accounts.

AI safety has also reached the government level: leaders of major technology companies signed a voluntary AI safety commitment at the White House.

Against this backdrop, the market for tools to control AI systems is developing: Cohesity offers to recover agents after failures, while Accenture offers independent model testing.

For investors, this means that opportunities in AI are not limited to model developers. As autonomous systems spread, demand is growing for solutions that can track their actions, identify errors, and reduce the associated risks.

We continue to monitor how the development of AI and new control tools affects the market and companies in the Raison portfolio. To discuss your investment portfolio, submit a request via the Raison bot and our experts will contact you.