Anthropic Surpasses OpenAI: A Financial Update

Summary of Anthropic Surpasses OpenAI: A Financial Update

by The Jaeden Schafer Podcast

14m•April 20, 2026
Listen to Episode

Overview of The Jaeden Schafer Podcast — Anthropic Surpasses OpenAI: A Financial Update

Host Jaeden Schafer reviews major recent shifts in the AI landscape: surprising growth in app launches, OpenAI’s product and exec shakeups (notably Sora’s shutdown), Cerebras filing for IPO and claiming a piece of OpenAI’s inference work, Stanford’s AI Index findings (US vs China performance gap and a big jump in agent capability), and a striking financial milestone — Anthropic’s reported $30B annualized revenue run rate surpassing OpenAI. The episode mixes data, market implications, and practical advice for people using AI today.

Key news highlights

  • App launches are booming: global Q1 app releases +60% YoY, iOS +80%; April releases (so far) up ~104% vs last year. Productivity and utilities are top categories.
  • OpenAI is shutting down Sora (text-to-video app): app closes April 26; API to follow in September. Three senior execs left around the same time, including product chief Kevin Weil and Sora head Bill Peebles.
  • Cerebras filed to go public: ~$23B valuation, reported $510M revenue (2025), $237M net income, AWS partnership, and a reportedly >$10B deal with OpenAI. CEO Andrew Feldman claims Cerebras took fast inference business at OpenAI away from NVIDIA.
  • Stanford AI Index: US vs China model-performance gap has shrunk to ~2.7% in favor of the U.S.; AI agents’ success on real software tasks rose from ~12% (a year ago) to ~66%.
  • Anthropic reportedly passed OpenAI in annualized revenue: Anthropic $30B run rate vs OpenAI $25B; Anthropic achieved this with meaningfully lower compute/training spend.

App-store resurgence — why it’s happening

  • No‑code / low-code “vibe coding” tools (Cloud Code, Replit, etc.) enable non-engineers to ship apps quickly.
  • Productivity and utility apps are leading downloads/launches.
  • Host’s view: rather than killing apps, AI is accelerating app creation and monetization.

Notable quote: Apple’s Greg Joswiak — “the rumors of the App Store's death have been greatly exaggerated.”

OpenAI & Sora shutdown: what it signals

  • Sora peaked (~1M users) but was costly (high compute burn); now being killed to refocus resources.
  • High-profile exits suggest internal strategy shifts toward coding and enterprise products.
  • Host interpretation: OpenAI narrowing focus to core enterprise/coding efforts, abandoning some consumer/side projects.

Cerebras IPO and chip competition

  • Background: Cerebras refiled IPO after pulling 2024 attempt (regulatory review over G42/Abu Dhabi investment). Raised large private rounds and now seeking mid-May IPO.
  • Financials & partnerships: 2025 revenue ~$510M; AWS deal to host Cerebras chips; reported >$10B agreement with OpenAI.
  • Strategic claim: Cerebras CEO says they captured fast inference workload at OpenAI that NVIDIA previously handled — potential crack in NVIDIA’s frontier-compute dominance.
  • Implication: the hardware/vendor landscape at the top end of AI compute could be shifting; worth monitoring for market and cost impacts.

Stanford AI Index — two headline takeaways

  • US vs China: performance gap on top model benchmarks narrowed to ~2–3% in favor of U.S. despite far smaller private investment in China.
  • Agents improvement: success rate on real, multi-step computer tasks went from ~12% → ~66% in one year.
  • Practical note: 66% accuracy is not safe for fully autonomous critical tasks, but good enough to scale narrow agentized workflows with human oversight.

Anthropic vs OpenAI — the revenue story and why it matters

  • Reported run rates: Anthropic $30B vs OpenAI $25B (annualized).
  • Growth velocity: Anthropic ~ $9B end of 2025 → $20B early March → $30B early April (rapid recent acceleration).
  • Customer and revenue mix:
    • ~80% of Anthropic revenue from businesses (enterprise-heavy).

    • 500 customers spending >$1M/year in February → >1,000 by April (doubling in ~2 months).

  • Compute/training efficiency: Anthropic reportedly spent ~4x less on training than OpenAI while generating higher revenue — implying much better unit economics if true.
  • Infrastructure commitments: Anthropic secured massive TPU/compute deals with Google (approx. 1 GW for 2026 plus 3.5 GW next-gen TPU capacity starting 2027).
  • Host’s perspective: being the enterprise/default coding model (Claude/Anthropic) is proving more lucrative than consumer scale for now; Anthropic seems to be winning enterprise dollars.

Practical takeaways / recommendations

  • Try AI agents for narrow, bounded tasks and supervise results — they’re much more capable than a year ago and can save significant time for data gathering, automation, and repetitive workflows.
  • If you’re an enterprise buyer, evaluate Anthropic/Claude for high-value automation and coding workflows — host reports outsized ROI.
  • Watch the Cerebras IPO and vendor compute deals for signs of shifting hardware economics (possible competition to NVIDIA at inference/training frontier).
  • If you build or launch apps: the App Store ecosystem is alive and benefiting from ease-of-shipping via modern AI-assisted dev tools.
  • Caution: agents at ~66% success need guardrails; don’t run mission-critical systems fully unsupervised.

Notable numbers and dates

  • App releases: Q1 2026 +60% YoY worldwide; iOS +80%; April releases +104% vs prior year window.
  • Sora shutdown: app closes April 26; API retired in September.
  • Cerebras: $23B valuation (reported), $510M revenue (2025), $237M net income; mid-May IPO target; >$10B reported deal with OpenAI.
  • Stanford AI Index: US lead over China ≈ 2.7%; agents success from 12% → 66% year-over-year.
  • Anthropic vs OpenAI run rates: Anthropic $30B vs OpenAI $25B (annualized); Anthropic used ~4x less training compute.

Host plugs / resources mentioned

  • AIbox.ai — host’s startup: unified access to 80+ AI models and an automation builder; priced at $8.99/month (promo mentioned).
  • AI Hustle School Community — paid/community resource for learning “vibe coding,” scaling businesses with AI.

Bottom line: the episode argues we’re in a fast-changing AI era where enterprise-focused models (Anthropic) and new hardware players (Cerebras) are reshaping market power, AI agents are becoming practically useful for many tasks, and the App Store is growing because AI makes app creation far easier.