DeepSeek
Scores out of 100 · Ranks within the 50-company cohort
Target date 2026-09-18
Target date 2027-09-18
Target date 2030-09-18
Our power thesis
We expect efficient models and developer adoption to expand its Compute and Interface power. Broad model distribution does not automatically give the producer permission control over downstream deployments.
Four Forces breakdown
Each force uses the same fixed 0–100 scale. The strongest force and breadth of control determine the Power Score.
| Force | 2026 | 2027 | 2030 |
|---|---|---|---|
| Compute | 65 | 75 | 85 |
| Interface | 60 | 70 | 80 |
| Alignment | 45 | 50 | 60 |
| Energy | 15 | 20 | 25 |
Our forecast
Across our 2026–2030 view, the Power Score rises from 55.6 to 73.8; the rank moves from #47 to #20.
Compute 65 → 85; Interface 60 → 80; Alignment 45 → 60; Energy 15 → 25.
The thesis above explains our call. Rank can fall even when power grows if other companies gain faster.
What would change our mind
Adoption fails to create durable infrastructure access or direct user relationships.
Research behind the judgment
DeepSeek’s current API documentation lists recent model and developer-platform updates, including September changes.
- DeepSeek API documentationCurrent September 2026 documentation · Accessed 2026-09-18
Sources inform the factual premises. Force scores, ranks, and forecasts are exmxc’s judgments as of this edition.