Key takeaways
Powered by lumidawealth.com
- Profit path: Anthropic targets breakeven in 2028; OpenAI projects ~$74B operating loss in 2028 and profitability by 2030.
- Burn vs. revenue (2025): OpenAI: $13B revenue / $9B burn. Anthropic: $4.2B revenue / ~$3B burn.
- Strategy: Anthropic skews to enterprise (≈80% of revenue), code and text; avoids compute-heavy image/video bets. OpenAI is pursuing broad consumer + enterprise with video (Sora), browser (Atlas), hardware, ads, e-commerce, robots.
- Capex/compute posture: OpenAI is committing up to $1.4T over 8 years and ~$100B in backup data-center capacity; Anthropic’s cost curve grows more in line with revenue.
- Valuations: OpenAI ~$500B; Anthropic ~$183B.
- Margin outlook: Anthropic’s model expands margins sooner; OpenAI accepts thinner near-term margins to chase scale and model leadership.
Why it matters
- Two dominant AI players are pursuing opposite risk profiles: a capital-intensive moonshot (OpenAI) vs. a disciplined, enterprise-first ramp (Anthropic). Investor tolerance for AI spend and the realized ROI on inference/training capex will decide which path wins.
By the numbers (from investor docs)
- 2025
- OpenAI: $13B sales, $9B cash burn (~70% of revenue).
- Anthropic: $4.2B sales, ~$3B cash burn (~70%).
- 2026
- Burn as % of revenue: Anthropic ~⅓; OpenAI ~57%.
- 2027
- Burn as % of revenue: Anthropic ~9%; OpenAI flat vs. 2026.
- 2028
- Anthropic: breakeven.
- OpenAI: ~$74B operating loss (~¾ of revenue).
- 2030
- OpenAI: profitability target.
Read on the approaches
- OpenAI
- Goal: multi-trillion-dollar platform; “risk of too little compute > too much.”
- Heavy stock-based comp to recruit talent; large reserved compute for research.
- New products: Sora (video), Atlas (browser), consumer hardware, ads, commerce, humanoids.
- Anthropic
- Focus: Claude for enterprises, coding, safety-oriented models.
- Avoids highest-cost modalities (video/image) for now.
- Cloud partners: Amazon, Google (vs. OpenAI’s Microsoft).
Risk checks
- Market risk: If AI revenue ramps slower than capex, OpenAI’s plan strains funding; if compute efficiency gains plateau, Anthropic’s lean approach looks prescient.
- Demand risk: Enterprise AI adoption velocity and unit economics of inference are pivotal.
- Regulatory/supply: Chips, energy, and data-center permitting could bottleneck both.
