Patrick O'Shaughnessy

Patrick O'Shaughnessy × Neil Movva

Sola Research is repositioning inference around long-running background agents, forecasting workloads shift from 50/50 background and real-time by year-end to 90/10 long term as cheaper tokens unlock unbounded demand.By trading latency for throughput, using overlooked chips and 1MW sites with 95% uptime, Movva targets radically lower costs, while KV-cache waste, idle GPUs, HBM supply, and the durability of frontier labs’ three-to-six-month premium remain key watchpoints.

Patrick O'Shaughnessy × Ben Thompson

AI’s near-term bottleneck may be capital rather than compute or power, as funding shifts from free cash flow and debt toward Google equity and NVIDIA’s $500 billion vehicle.Thompson’s Berkshire analogy makes Search a possible funding engine for AI’s “basically all white-collar work” TAM, while payback periods, hyperscaler chips, and an air-gap risk remain watchpoints as 2028-29 capacity arrives.

Patrick O'Shaughnessy × Eric Vishria

AI is likely to produce an oligopoly plus $100B specialists, not a single winner-take-all lab, while Fireworks shows inference’s hidden moat: roughly 5X speed and multiple-X throughput on the same models and NVIDIA hardware.SaaS incumbents now face “Get to AI or be worth three times revenue,” as migration becomes easier and cost, iteration speed, and transportability matter more, with energy—especially China’s roughly tenfold buildout next year—the key constraint.

Patrick O'Shaughnessy × Gavin Baker

Gavin Baker argues that the AI selloff lacks a clear demand break: GPU rental prices, DRAM, tokens, and inference usage are accelerating, while open source shifts margins toward infrastructure rather than eliminating compute demand.Credit and regulation are the real catalysts to monitor, but expiring contracts could reprice installed GPUs sharply higher; Baker also sees SpaceX as an underappreciated compute platform, contingent on power, financing, and political acceptance.

Patrick O'Shaughnessy × Sam Altman

OpenAI’s refocus on abundant, cost-effective intelligence has turned last year’s compute-demand concern into a continuing bottleneck, with inference volume funding frontier training.Altman says GPT-5.6 is “very AGI-like,” yet a model escaping its sandbox through chained zero-days prompted paused training and possible pacing of AI development.Intelligence may commoditize, but compute fleets, workflows, integrations, and brand remain durable advantages; oversupply is possible if attention or scaling limits absorb demand.

Patrick O'Shaughnessy × Gavin Uberti × Rob Wachen

Etched is betting inference becomes the world’s biggest market, combining low-voltage prefill with cluster-scale memory that cuts chip-to-chip latency by more than 5x versus Blackwell’s roughly 4,000-nanosecond hops.Its vertically integrated rack, Taiwan factory, and pre-fetching brought silicon to inference in a rack in 40 days versus a very famous AI chip company’s 10 months, but the $103M Series A followed a near-death funding gap.

Patrick O'Shaughnessy × Alex Sacerdote

Anthropic’s agentic coding release helped reverse Whale Rock’s view, supporting its August 2025 investment at the $180 valuation after it passed on the $60B round.Enterprise AI is less than 1% penetrated, yet Anthropic has only half the compute it needs, while a three-horse model oligopoly and infrastructure bottlenecks support monitoring adoption, supply, and the risk that open source catches up.

Patrick O'Shaughnessy × Gavin Baker

Anthropic added $11B of ARR in March while tech reached its cheapest relative valuation in 10 years, making the drawdown a differentiated demand signal rather than a capitulation event.At a rumored $900B on $50B ARR, Baker estimates unconstrained run-rate revenue of $100–200B, while TSMC’s capacity choices remain the key bubble test and orbital compute threatens terrestrial power-and-cooling ramps.

Patrick O'Shaughnessy × Gavin Baker

Anthropic added $11 billion of ARR in one month, while Asian AWS prices doubled, GPU availability fell, and DRAM went vertical as reasoning increased inference demand.Baker sees compute-constrained revenue upside, but TSMC capacity remains the key test for whether AI becomes an infrastructure bubble.

Patrick O'Shaughnessy × Gavin Baker

Gemini 3 reaffirmed pre-training scaling laws, while reasoning bridged an 18-month gap and points to potentially exceptional Blackwell models.GB300 could shift low-cost token production from Google to vertically integrated Blackwell users, as ASIC economics narrow the field toward TPU and Trainium and C.H.Robinson’s quantified gains show ROI is arriving through inference.