Cohere is raising billions of dollars to expand its AI infrastructure, but the company does not plan to build every data center itself, and its chief AI officer says partnerships with neocloud providers remain part of its compute strategy.
Cohere · AI infrastructure
Cohere is raising funds for compute, but will not build every data center itself
The new capital targets more computing capacity to train and run its models. Its Chief AI Officer says partnerships with neocloud providers remain part of the plan.
How the compute strategy fits together
Step 1
New funding
Aimed at expanding compute
→
Step 2
Compute capacity
Needed to train and run models
→
Source A
Self-built data centers
Direct control, heavy upfront spend
+
Source B
Neocloud partners
Faster to add, flexible spending
The model is mixed: the company is not replacing outside partners with an entirely self-built network.
Why neocloud partners help
Capacity can be added more quickly
Spending can adjust as demand changes
Avoids long lead times for power and hardware
Draws on providers that specialize in GPU computing
The trade-off
Core input depends on third parties
Less direct control over infrastructure
Cost structure and speed tied to partner terms
Self-built capacity requires heavy upfront spending and operating expertise
What remains unclear
The split between owned and partner capacity will determine how much infrastructure Cohere controls directly.
Funding size
Not specified
Neocloud partners
Names not disclosed
Capacity split
Self-built vs partner not divided
Enterprise customers and investors will watch how this balance affects reliable compute at scale, cost structure, and development speed.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...