CNBC frames OpenAI and Anthropic as IPO candidates being priced like enduring AI monopolies, then argues that cheaper, increasingly capable open-source models—especially from China—are compressing pricing power and weakening the moat behind those valuations. The interview with Cohere CEO Aidan Gomez adds a more nuanced counterpoint: enterprise demand remains strong, but the market is shifting toward smaller, more efficient, security-focused models deployed on-prem or in private environments.
Watch on YouTube ›Get the market thesis, key claims, assets, contradictions, and follow-up questions from any financial video — then unlock a version personalized to your portfolio, watchlist, and favorite speakers.
The video’s core thesis is that OpenAI and Anthropic are being marketed to public investors as trillion-dollar, decades-long pricing-power franchises, but that pitch is getting challenged by the reality of cheaper AI. CNBC argues that Chinese open-source models have rapidly improved, are winning usage share, and can do many enterprise tasks at a fraction of the cost of premium frontier models. The segment repeatedly emphasizes that the “moat” sold to IPO investors may be shrinking in real time, especially if customers begin optimizing for cost and task-specific performance rather than simply chasing the biggest model. A major supporting point is the price/performance comparison. …
Tactically, the setup is bearish on premium AI valuations: the market may start discounting OpenAI and Anthropic if cheaper models keep winning share and enterprise buyers get more cost-conscious. Short term, headlines around benchmark parity, usage shifts, or IPO timing could drive volatility in the whole AI complex.
Over the next few months, the likely path is bifurcation rather than collapse: frontier labs can still grow, but efficiency-focused and security-focused models may take a larger share of enterprise budgets. Confirmation would come from continued adoption of smaller or private deployments; invalidation would be broad willingness to pay up for frontier models despite cheaper substitutes.
Structurally, AI looks less like a winner-take-all frontier-model market and more like a layered ecosystem where trust, deployment, and cost efficiency determine margins. If that holds, the public-market premium may shift away from raw model capability toward the infrastructure and distribution stack that sits around it.
OpenAI and Anthropic are being pitched to investors at valuations north of $800 billion, with the market soon to test whether that pricing is justified.
The intro states both companies are courting investors for IPOs above $800 billion and asks whether they are worth $1 trillion each.
Chinese open-source models are eroding the moat of frontier AI labs by matching capabilities at lower cost and attracting usage share.
The segment argues Chinese models are cheaper, competitive on benchmarks, and gaining usage share on OpenRouter.
Claude Opus can cost roughly nine times more than the cheapest Chinese alternative for similar work.
This is the segment's main quantitative price comparison supporting the commoditization argument.
Is premium AI, frontier AI, still worth it despite the cost?
Aidan Gomez says yes — the demand shows people are willing to pay, and companies' level of CapEx spend and reduction in free cash flow demonstrates they can see future demand ramping. He notes the core bottlenecks to enterprise AI adoption are cost and security.
Talk a bit about Cohere's business model — how are you thinking about cost payoff and benefits for the enterprise?
Cohere builds models from scratch exclusively for the enterprise, focusing on high-security settings like grid operators, financial services, telecom, and government. They offer very secure deployments — on prem or completely airgapped, even inside submarines — which gives them a unique value proposition. Their models must run on just 2-4 GPUs, ruling out massive frontier models.
Is that cost calculation for enterprises changing as open source models get very close to the frontier and the lag closes?
Yes, cost control and the compute bottleneck are driving a shift. There's simply not enough compute to support using massive models, so efficient small models that are good enough are needed. Everyone raced to adopt AI at all costs, but now CFOs will look at expenses and try to optimize — though they won't pull back, just find smaller, more efficient models.
Unlock the full claims, asset map, scores, related transcripts, follow-up questions, and AI chat — shaped around your portfolio, watchlist, favorite speakers, and risks.
