Current Compute Footprint and Funding Ineffable Intelligence launched with an exceptionally strong financial profile, raising a $1.1B seed round at a $5.1B valuation in April 2026 with backing from Alphabet/Google, Nvidia, Sequoia, and the UK Sovereign AI unit 44 sources. However, the company currently has no publicly quantified compute commitments. Its primary infrastructure consists of an Nvidia engineering partnership for RL co-design blogs.nvidia.com and a Google Cloud agreement to deploy "one of the largest clusters" of A5X instances powered by Vera Rubin NVL72 GPUs 3 sources. While the first of these instances was live by July 2026 blogs.nvidia.com, neither deal discloses a GPU count, megawatt target, or contract value. Furthermore, access to UK national supercomputers via the Sovereign AI fund is excluded from these estimates, as these are shared resources rather than dedicated, exclusive-use capacity sovereignai.gov.uk.

The Arithmetic of Scale and Capital Estimating capacity requires mapping Vera Rubin deployments to power and cost equivalents. Based on standard hardware facility overheads, a Rubin-class GPU requires roughly 2.0 to 3.0 kW of all-in power epoch.ai. At this ratio, a massive deployment of 100,000 Rubin GPUs translates to roughly 300 MW (0.3 GW). Financially, renting 100 MW of frontier GPUs costs approximately $0.7B to $1B per year. Consequently, Ineffable's entire $1.1B seed round can only sustain on the order of 100 MW for a single year. Scaling to a gigawatt or more by the end of 2028 would require capital expenditures or rental agreements exceeding $10B—a monumental leap requiring unprecedented follow-on funding or massive vendor financing.

Trajectory to 2028 (Median Assessment) Our median estimate of ~0.28 GW reflects a highly successful scale-up constrained by the sheer cost of infrastructure. A capacity of nearly 300 MW represents approximately 70,000 to 100,000 dedicated Rubin or next-generation GPUs. We expect Ineffable to reach this tier by securing one or two fully dedicated, exclusive-use mega-clusters on Google Cloud, funded by a substantial follow-on financing round in 2027. Reaching this scale aligns with the ambition of their RL-driven "superlearner" mission cnbc.com while respecting the financial realities that separate well-funded startups from hyperscalers.

Key Uncertainties and Tail Risks The variance in the forecast is heavily tied to future capital availability and strict interpretations of the resolution criteria. The upper tail (p90 of ~1.37 GW) accounts for the possibility that Alphabet or Nvidia structurally underwrites a massive, multi-gigawatt buildout—effectively turning Ineffable into a flagship European AI customer and bundling equity with computing power ahead of visible funding. Conversely, the lower tail (p10 of ~0.05 GW) reflects the risk that the Google Cloud partnership acts as a rolling cloud reservation rather than a strict, exclusive-use contract, which could fall short of the required resolution criteria. Downside risk also exists if Ineffable's heavy reliance on reinforcement learning environments ultimately requires less centralized training capacity than traditional frontier LLM labs, keeping their live, dedicated footprint in the tens of megawatts.