Cheap AI could derail OpenAI and Anthropic IPOs. DeepSeek matching benchmarks, enterprise costs surging, moats eroding. - genuine question

[Lucy05]

CNBC published a detailed piece this week examining how the commoditisation of frontier AI capability threatens the IPO valuations both OpenAI and Anthropic are pursuing, both projected north of 800 billion dollars. DeepSeek's next-generation model preview matches or nearly matches the latest from all three major US labs on coding, agentic, and knowledge benchmarks. Models from Moonshot, Xiaomi, and Zhipu have shipped at similar capability levels in the past four months.

Meta, Shopify, Spotify, and Pinterest all flagged rising AI inference costs as a margin drag in this earnings season. 45 percent of companies surveyed spent more than 100,000 dollars monthly on AI in 2025, up from 20 percent the year before. The gap between what enterprise customers are paying and what open-weight alternatives cost is becoming harder to explain.

Cheap AI could derail OpenAI and Anthropic's IPOs

https://www.cnbc.com/2026/05/20/cheap-ai-could-derail-openai-and-anthropics-ipos.html

[One-One-Five]

The benchmark convergence between DeepSeek and US frontier models is the structural problem that no amount of brand marketing solves. If the capability is equivalent the pricing premium requires a different justification

[Red Builder]

Meta's open-weight Llama models are the domestic US version of this problem. OpenAI and Anthropic are competing against open-weight alternatives that their primary competitor is giving away

[Candle]

The 800 billion dollar IPO valuation requires sustained pricing power in the enterprise segment. Every quarter that DeepSeek maintains benchmark parity reduces the multiple investors will accept

[FridayFeeling]

Inference costs being flagged as a margin drag by multiple public companies in the same earnings season is the demand-side signal that something has to change about pricing. Either the cost comes down or usage adjusts

[Mark7]

The CNBC piece framing this as could derail is conservative. The more accurate framing is that this is the primary risk that IPO investors need to model explicitly and many are not

[Freya]

The US government's CHIPS Act quantum investment and export controls on AI chips are partly about preventing exactly this dynamic in the next technology cycle. The AI cycle lesson is being applied to quantum while there is still time