OpenAI, Google Turn AI Speed Into a Selling Point

Generative AI competition has largely centered on model capability and price, but latency is becoming a commercial differentiator as companies deploy customer-service bots, coding agents and financial-analysis tools. Faster inference lets software complete more steps while an event is still unfolding, improving productivity and user engagement. OpenAI underscored the shift on Jan. 14 with a multiyear Cerebras agreement for 750 megawatts of low-latency computing capacity, a deal reported at more than $10 billion and scheduled to roll out in stages through 2028.

On Aug. 13, OpenAI previewed Ultrafast, an API service tier that runs GPT-5.6 Sol on Cerebras chips at as many as 750 output tokens a second, up to 14 times faster than Standard processing. Access is initially limited to selected customers and pricing has not been disclosed. Google released Gemini 3.7 Flash the same day, charging an introductory $0.75 per million input tokens and $3.75 per million output tokens through Dec. 31; those rates double on Jan. 1, 2027.