Bristol Myers Squibb takes the lead seat at Pathos AI

Covers roughly 5am ET Friday, Oct 23 through early Monday, Oct 26. It was a quiet weekend: four AI rounds in the window cleared our checks. We also include seven Oct 19–22 rounds that earlier issues missed (marked "catch-up"). Every round below is confirmed by the company or its lead investor. Figures companies report about themselves are labeled company-reported.

Pathos AI, a New York company that uses AI to choose and steer its own cancer-drug programs, closed a $237M Series E at a post-money valuation of about $3B. Bristol Myers Squibb led, and existing investors also took part (Pathos · SiliconANGLE). The equity came with a commercial agreement. BMS gets non-exclusive access to Pathos's AI platform, Foundry, across its wider R&D portfolio, plus three years of exclusive rights to use it in support of its neuroscience programs. Pathos says the collaborations begin with Alzheimer’s and other neurological conditions. Separately, Pathos will pay Tempus AI $55M over five years to license a de-identified dataset Tempus is building that will cover several hundred thousand people with neurological disorders (Pathos collaboration release). Pathos says Foundry uses thousands of task-specific AI agents to keep each program's clinical, genomic and competitive data current and to recommend the next development step (company-reported). It now has four clinical-stage assets. The top use of the money is pushing JSKN016, a TROP2/HER3 bispecific antibody-drug conjugate, toward registrational studies. Why it matters for investors: The lead investor is also a customer. BMS's check arrives bundled with platform access and a three-year neuroscience exclusive, so part of what the other shareholders own is a platform whose neuroscience use is reserved for one drugmaker for three years. On the headline numbers, the new money is roughly 8% of the post-money valuation (our arithmetic), light dilution for a company with four clinical programs. But this is a drug developer, not a software company: Pathos owns its assets, so returns depend on trial readouts, with AI as the claimed edge in picking and steering them. When an AI-drug company brings in a pharma lead, ask exactly which rights the strategic received and whether they limit who else can partner with the company or buy it.

Historical lens. Pharma has used a funding round to get closer to a data platform before. In January 2016 Roche led Flatiron Health's $175M Series C. In the same announcement, Flatiron said it would enter a multi-year, non-exclusive agreement under which Roche would buy several of its life-science offerings, and that Roche Pharmaceuticals COO Daniel O'Day would join its board (Flatiron). Two years later, Roche signed a deal to buy all of Flatiron, building on an existing 12.6% stake, for a $1.9B payment on a fully diluted basis. It said Flatiron would remain a separate legal entity and keep serving its other partners (Roche). Roche had followed a longer version of this path with Genentech: a 1979 partnership on interferons, the purchase of 60% of Genentech in 1990, and Genentech joining the Roche Group in 2009 (Genentech). Our interpretation, not established fact: BMS's package of equity, platform access and a limited exclusive looks like the first step of that sequence. Whether it leads anywhere further is unknown. For Pathos's other investors, the open question is whether the strategic lead ends up as the likeliest buyer, and so the party that sets the exit price.

Higgsfield, a San Francisco platform on which brands, agencies, studios and creators make images and video with AI, extended its Series B with new money from TCV and NVIDIA. It did not disclose the amount or a valuation. DST Global led the original Series B in August (Higgsfield / PR Newswire · Bloomberg). The company says it passed $1B in annualized revenue in September, 18 months after launch, that its enterprise business is more than doubling month on month, and that it has more than 40M users. It also says it runs 150 million generations a month, totaling 12 trillion tokens (all company-reported). Higgsfield says it has run on NVIDIA hardware at every stage and is now expanding to NVIDIA's Vera Rubin generation. Why it matters for investors: An extension with no number attached tells you who wanted in, not at what price, so read it as a signal rather than a new mark. The harder question is gross margin. Higgsfield describes its product as leading video, image and audio models brought together with its own creative technology, which raises the question of how much of each dollar of revenue goes to model providers and compute. The company cites industry benchmarks putting generative video's yield at $30–$150 per million equivalent tokens, against $0.15–$0.60 for text. That spread is the bull case for visual-AI applications. The bear case is that the model suppliers capture it. In diligence, ask how "annualized revenue" is defined and what share of generations run on the company's own models.

Blackwall makes traffic-security software that sits within the infrastructure of web-hosting companies. It raised a $100M Series C at a $2.2B valuation. AVP led, and existing investors Dawn Capital, MMC Ventures and Tera Ventures also took part (Blackwall / Business Wire · TechCrunch). Its products span bot mitigation, DDoS protection, a web application firewall, load balancing, server-side encryption and management of AI-agent traffic. Partners can resell the modules under their own brands and set their own tiers. Blackwall's executive chairman told TechCrunch the valuation is four times what it was at the company's 2025 round. The Estonia-based company (per TechCrunch) calls itself Europe's most valuable cybersecurity scaleup (company-reported). Why it matters for investors: Every agent that browses, scrapes or shops for a user is traffic that a website must sort into welcome or unwelcome. Blackwall's bet is on distribution. It doesn't sell to websites in competition with Cloudflare or Akamai. It sells through the hosting providers that already own the customer relationship, under a revenue-share arrangement (per TechCrunch). That channel can scale cheaply, but it makes the partners gatekeepers. TechCrunch reports that CEO Nikita Rozenberg expects sites will eventually need controls over both agent access and agent payments. For investors, the question is whether sitting at that checkpoint earns a toll on agent commerce or stays a security cost that hosts negotiate down.

Round notes

Not investment advice. The AI Cap Table is published for information only and is not investment, legal, or tax advice. Nothing in this newsletter is an offer or solicitation to buy or sell any security. Funding data comes from public sources we cite and verify, but errors are possible; do your own due diligence and consult a qualified adviser before making any investment decision.