The DeepSeek shock: AI capex assumptions get repriced
Backfilled context carries a typed causal-confidence label and never claims a confirmed driver without human-attached evidence — a single reviewed reference establishes chronology, not causation.
What happened
On 27 January 2025, claims that Chinese startup DeepSeek had trained a competitive AI model at a fraction of assumed cost triggered a broad repricing of AI-infrastructure stocks. NVIDIA fell roughly 17% in one session — the largest single-day market-value loss for any company up to that point — and AI-linked names fell across the board.
What was known then
At the time, DeepSeek's model was publicly available and its performance claims were being rapidly tested, but its true training cost, hardware access and reproducibility were all disputed. The selloff traded on a question, not a settled fact.
Market reaction
AI-semiconductor and data-centre-linked stocks fell sharply for a session, with partial recoveries in the following days as analysts debated how much compute demand the cheaper-training story actually threatened.
What was uncertain
Whether cheaper training genuinely reduced long-run chip demand or increased it (the efficiency-rebound argument), what hardware DeepSeek had actually used, and whether US export controls had failed or worked.
What changed afterward
AI capex guidance from the major platforms stayed elevated in subsequent quarters, and the episode settled into the reference case for how fragile consensus AI assumptions can be against a single credible challenge.
Why it still matters
AI-infrastructure names remain a core coverage area; every earnings cycle since has been read partly through the question this episode raised — what happens to the AI trade when its cost assumptions move.
Sources
Primary and official references reviewed for this entry.
◦ NVIDIA Investor Relations (company)
◦ U.S. Securities and Exchange Commission (official)
Latest developments
Current coverage that genuinely overlaps this episode — nothing is linked for the sake of linking.
NVIDIA's quarterly filing for the period ended April 26, 2026 reports revenue of $81.6 billion, up from $44.1 billion a year earlier. Data centre revenue was $75.2 billion, up 92% year on year, with hyperscalers about half of it. Net income was $58.3 billion with basic earnings per share of $2.40. No data centre Hopper products were shipped to China, against $4.6 billion a year earlier.
SAP announced on July 17 that it completed the acquisition of Prior Labs, which it describes as the pioneer of Tabular Foundation Models. Prior Labs will continue operating independently, with SAP committing more than €1 billion over the next four years to scale it into a frontier AI lab for structured business data.
SAP reported second-quarter 2026 results with current cloud backlog of €22.9 billion, up 27% (26% at constant currencies), cloud revenue up 22% (24% cc) and total revenue up 9% (11% cc). IFRS operating profit rose 8%. The company updated its 2026 non-IFRS operating profit outlook to reflect dilution from the Dremio and Prior Labs acquisitions.
Samsung Electronics reported second-quarter 2026 consolidated revenue of KRW 171.5 trillion — an all-time high, up 28% on the prior quarter — and record operating profit of KRW 89.5 trillion. Earnings per share rose 52% to KRW 10,849. The Device Solutions division, at KRW 127.5 trillion revenue and KRW 89.2 trillion operating profit, drove the quarter as the memory business set records on AI server demand.
For market-reference purposes only. Trading/investment decisions remain the user's responsibility.