Baseten closed a $1.5B Series F led by Altimeter Capital, Conviction, and Spark Capital at a $13B valuation — up from $5B in January — as revenue grew approximately 20x year-over-year and the platform hit one billion inference calls per day across 87 clusters spanning 18 clouds. The valuation trajectory positions multi-cloud inference orchestration as a defensible infrastructure position, not a hosting commodity, and the enterprise software layer wrapping GPU access is where the margin lives.
For a direct account of how CEO Tuhin Srivastava thinks about the inference crunch, GPU contracting dynamics, and why the software layer retains customers at scale, we enjoyed his May 2026 interview on No Priors.
Groq confirmed a $650M growth round led by Disruptive and Infinitum on Monday, with capital earmarked for inference capacity expansion toward 200 megawatts by 2027. CEO Adam Winter and CFO Matt Eng are leading the company after founder Jonathan Ross departed to Nvidia in December under the $20B LPU licensing deal; new hires include COO Alan Rice (previously at xAI and Meta), CTO Sinclair Schuller, and CPO Rakesh Malhotra, both joining in July. The company operates 13 data centers across four continents serving more than five million developers and processes trillions of tokens weekly — the bet now is whether LPU-based inference speed holds a defensible neocloud position as Nvidia's own LPX product, built partly on Groq's licensed architecture, enters the market alongside it.
The open-source AI lab founded by former Google DeepMind researchers Misha Laskin and Ioannis Antonoglou signed a compute lease with SpaceX worth up to $6.3B — $150M per month starting July 1 through 2029 — for Nvidia GB300 access at SpaceX's Colossus 2 data center in Memphis. The deal is SpaceX's third external compute lease (following Anthropic at $1.25B/month and Google at $920 million/month), bringing committed Colossus monthly compute revenue to $2.32B across three customers. Reflection, which competes at the frontier with closed-model labs on an open-weight strategy, is now training and serving on the same physical infrastructure as its competitors — compute access as parity rather than competitive differentiation, and SpaceX's Colossus as the neutral venue for a contested AI market.
SpaceX fell 16.4% on Monday — its worst session since the June 12 Nasdaq debut — after announcing a $20B bond offering, its first major debt issuance. The decline is a balance-sheet reaction, not a response to the Reflection AI compute deal; the market is pricing leverage risk on a company that went public ten days ago. With $2.32B in committed monthly Colossus compute revenue now on the books, how quickly that cash flow covers the cost of capital is the operative question for the next phase of the thesis.
CoreWeave fell on the effective date of its Nasdaq-100 inclusion — alongside Nebius, Astera Labs, Rocket Lab, and Teradyne, all of which also declined — as active traders who front-ran the mechanical rebalancing exited at the close. More than $800B in passive AUM now tracks these names.
Micron reports fiscal Q3 results after market close. The company guided for approximately $33.5B in revenue at roughly 81% gross margin. HBM4 capacity is fully booked through early 2027; the key language to watch is ASP trajectory, supply allocation across HBM tiers, and any updated guidance on the pace of the memory supercycle.
Nvidia holds its 2026 Annual Meeting of Stockholders virtually at 9am PT. Blackwell Ultra production ramp and Vera architecture timing are the primary items investors are tracking; any updated language on data center demand or the inference product roadmap could move the stock.