Anthropic's confidential S-1 filing for a fall 2026 public offering frames what is shaping up to be the largest pure technology IPO in US market history. The company closed a $65 billion financing round at $965b, and the capital markets absorbed the raise as effectively a rounding error, a signal that institutional demand for frontier AI equity is operating at a different order of magnitude than prior technology cycles. Annualized revenue has reached $44 billion, up from $9 billion at the close of last year, a growth rate that places the $965 billion valuation on a forward multiple that begins to look less stretched with each passing quarter. OpenAI has filed its own confidential S-1 and was last marked at $852 billion following a $122 billion first-quarter capital raise, putting both companies on a parallel trajectory toward H2 2026 listings. The three-IPO supply question, whether the simultaneous pipeline of SpaceX, OpenAI, and Anthropic overwhelms available capital, appears misframed; the global equity market is a multi-trillion dollar pool that has already absorbed tens of billions from each of these companies in private rounds without pricing friction. Both Anthropic and OpenAI are expected to test trillion-dollar-plus valuations at pricing.
Baseten's move to raise $1 billion at an $11 billion valuation, a 120 percent step-up from the $5 billion mark set in January, is one of the cleaner data points available on where the AI infrastructure stack is repricing in real time. The business sits between GPU-bearing data centers and the software companies building on top of AI models, ensuring compute resources are continuously utilized on the supply side while delivering the lowest available cost-per-token to application developers on the demand side. That positioning eliminates the need for developers to sign multi-year hyperscaler contracts, an advantage that is especially valuable to earlier-stage AI companies whose compute requirements are still accelerating and unpredictable. Revenue is targeting a $1 billion annualized run rate by year-end, a target anchored by a revenue triple in a single quarter, the type of acceleration that historically reflects a product finding genuine enterprise pull rather than a sales-cycle artifact. The implied forward multiple of approximately 11x revenue is not obviously expensive given that each new AI workload going into production is additive to Baseten's revenue base, and the volume of workloads is still in its early innings. Decart AI, a video generation platform, used Amazon Trainium 3 chips to compress video workflow compute costs from $100 per hour to $0.25 per hour while achieving 8x faster inference throughput, illustrating the scale of the optimization opportunity that Baseten's architecture is positioned to capture. Every enterprise token burned across any workflow is a revenue event for companies operating in this middleware layer, and the demand curve for inference is compounding rather than plateauing.
The most underappreciated constraint on enterprise AI adoption is not model quality; it is the absence of internal implementation capability, and a rigorous MIT study finding that 95 percent of corporate AI initiatives fail to generate measurable profit confirms the pattern. Forward deployed engineers are the structural response to this gap, operating as AI-specialized consultants who embed within individual business units, map the full inventory of repetitive workflows, and build automation layers that typically eliminate 75 percent of manual task volume per department. The role exists because most enterprises understand that AI works and that adoption is competitively necessary, but they do not have the engineering talent in-house to design, deploy, and optimize agent-based systems at production scale. Glean addresses a critical leverage point in this workflow: by tuning prompt architecture and constraining context windows to the precise data needed for a given task, the company reduces token consumption by up to 90 percent while preserving output quality, which is the difference between an AI initiative that produces a compelling ROI case and one that stalls in the pilot phase. The longer-term structural shift underway is the diffusion of software development capability into every business unit, a generational transition analogous to the spread of Excel literacy across finance and operations teams, where self-service handles the standard work and specialists handle the complex edge cases. Anthropic's developer community guidance from this week adds a forward-looking dimension: design automation workflows 6 to 12 months in advance under the assumption that the next generation model will natively resolve whatever you are coding today, a posture that demands exactly the planning and deployment expertise that Glean and the broader FDE cohort are building as a durable moat.