Meta and Anthropic at the same valuation: one is bound to be vastly wrong
Meta's current valuation sits under $2 trillion; Anthropic, heading toward IPO, prices at comparable levels. One builds profit from advertising; the other from AI models alone. The author argues one of these valuations will face substantial revision.
The video won't play here. Listen to the audio instead:
The argument · timestamps estimated from transcript position
Muse Spark Open halves the open-source gap between US and China
Muse Spark Open's performance sits between GLM-5.2 (stronger) and DeepSeek-V4.1-Flash (weaker), still trailing GLM-5.3 and Kimi K3. The author judges that without Muse Spark Open and Reflection's Beam, US open-source models would be comprehensively outpaced by 20+ Chinese alternatives. This release shifts the US's strongest model from ‘performance-irrelevant’ to ‘behind but visible’—cutting the US-China open-source gap roughly in half.
Custom license makes Muse Spark Open barely open-source
The license reproduces nearly every restriction the open-source community has criticized: requiring ‘Built with Muse’ attribution in products; mandating separate contracts for companies earning over $20M monthly; including an Acceptable Use Policy extending restrictions to model outputs and derivatives, creating legal gray area around synthetic data generation. The author calls this complex, non-standard license an unnecessary strategic risk for a lab desperate to recover adoption.
Meta is the only US company spanning open and closed frontiers
Days before Muse Spark Open's launch, Meta announced closed-source models Muse Wonder and Muse Wonder Full, which rank just behind Anthropic's Opus 5.5 on standard benchmarks and match GPT-6-Astra and Gemini-4-Argon. Meta alone among US companies fields both frontier-leading open and closed models; Alibaba's Qwen follows a similar dual path but leans open. This comes barely a year after Zuckerberg's widely criticized ‘Personal Superintelligence’ post.
Flat management structure was Meta's key to AI recovery
Building models requires capital, compute, data, and talent—but organizing those resources is hardest. Alex Wang oversees over 200 direct reports in an exceptionally flat structure yet ships steady model and product improvements. By contrast, Gemini is often described as a place where things struggle to ship; Llama's earlier failures were blamed on management deciding by politics rather than performance; xAI may face talent churn and insufficient organization to sustain top-tier models long-term.
Anthropic leads OpenAI in overall capability ranking
The author's composite ranking places Anthropic above OpenAI, Meta and Gemini at parity, xAI with China's open models, and US open models at the bottom. Anthropic's organization is clearest and most methodical, establishing a lead. OpenAI shows more volatility but consistently holds the frontier in deployed models while leading in published breakthroughs—from Navier-Stokes to o1's reasoning-time scaling.
US frontier players have narrowed to five contenders
The author sees the prediction ‘frontier models will become too expensive for anyone but megacap’ playing out in practice: only Anthropic, OpenAI, Meta, Gemini, and xAI can truly compete. If Microsoft begins releasing proprietary models, its scale economics and ability to train on OpenAI IP directly could crack this exclusive list.
Third-tier labs increasingly struggle to reach the frontier
Mistral, Nemotron, Reflection, Arcee and other ‘third-tier’ players' best models increasingly struggle to move the frontier. They still capture meaningful adoption—enough to sustain a business—but building a true frontier model now costs billions per flagship. Only giants absorb this scale. The result: top-tier positions calcify.
Meta and Anthropic valuations will not both survive unrevised
Meta currently values under $2 trillion; Anthropic, heading to IPO, prices at similar levels. The author argues both companies start at roughly the same long-term potential, yet Meta owns a cash-generating core business. As Meta and other challengers stay in play, AI-related IPO valuations will remain pressured—before even pricing in safety risk. The author poses: what will Anthropic's public market cap be by November 2027? By 2028?
In their own words · checked verbatim
Without these two models, there would be 20+ open-weight models built and released by Chinese companies better than the best Western models.
If anything will hold Muse Spark Open back, it’ll be the specific license they designed for it.
Meta is the only American company with frontier open and closed models.
Alex Wang has over 200 direct reports — i.e. it’s a very flat org with minimal management — and they’re releasing very consistent model and product improvements.
One of the two valuations needs to correct massively.
Yes, evaluations are a good proxy, but why do the leading labs like Anthropic go from hit models like Claude Fable 5 to total duds in Claude Opus 5 and then back to hits with Opus 5.5?
Figures
| Muse Spark Open size | 816B total parameters / 42B active parameters, 1M token context window | 0:00 |
| Alex Wang direct reports | Over 200 | 5:25 |
Glossary
- RSI
- Recursive self-improvement: a posited path to exponential capability gains through iterative self-refinement.
- tier 1/2/3 labs
- Informal industry ranking of AI labs by overall capability; tier 1 is strongest.
How to listen
Investors and engineers tracking the open-source LLM ecosystem, competitive dynamics among AI labs, and Series funding valuations.
Listeners uninterested in licensing detail can skip the opening rundown of license terms.