Firmus secures OpenAI anchor deal, pushes contracted AI capacity past 900 MW

Deal Overview

On September 8, 2026, Australian AI‑infrastructure firm Firmus announced a multi‑year agreement with OpenAI to provide compute capacity from two new data‑centre sites in Malaysia. The contract designates OpenAI as an anchor customer, guaranteeing a baseline of usage for the facilities once they become operational.

Why OpenAI as Anchor Customer Matters

OpenAI’s position as the leading provider of generative‑AI services gives the Firmus deal immediate credibility. An anchor customer arrangement signals that the capacity will be monetised at market‑rate pricing, reducing the revenue uncertainty that typically accompanies early‑stage data‑centre builds. For investors, the OpenAI link also suggests a downstream pipeline of AI workloads that are likely to expand as OpenAI rolls out newer model families.

Contracted Capacity Reaches Over 900 MW

The OpenAI agreement lifts Firmus’s total contracted capacity across all customers to **more than 900 megawatts**, a figure that now spans a mix of enterprise, research, and OpenAI workloads. A capacity pool of this size is a material benchmark in the AI‑infrastructure sector, where most single-site projects are an order of magnitude smaller.

Pre‑Selling Capacity Before Power Is Online

The strategy of selling compute capacity ahead of the actual power delivery is a hallmark of the AI‑infrastructure boom. By locking in demand now, Firmus can lock in financing terms that reflect a future cash‑flow profile, rather than relying on speculative take‑rate assumptions. The approach also mitigates the risk of under‑utilisation once the Malaysian sites connect to the grid, because the revenue stream is already pledged.

From a pre‑IPO standpoint, this pre‑sell model improves the clarity of the company’s forward‑looking metrics. Analysts can model revenue based on firm commitments, rather than a blend of signed letters of intent and speculative pipeline projects. It also provides a tangible narrative for investors: a substantial portion of the forthcoming power budget is already allocated, supporting a smoother transition to profitability after the sites become operational.

Investor Base and Valuation

Firmus counts a distinguished set of backers: Nvidia, the chip‑design leader that has a strategic interest in AI‑compute capacity; Jane Street, a quantitative trading firm with deep expertise in technology‑focused investments; Blackstone funds, bringing heavyweight private‑equity discipline; and Coatue, a technology‑focused growth investor. The latest fundraising round valued Firmus at **more than $10.5 billion**.

This valuation reflects both the market‑wide scarcity of large‑scale AI‑compute infrastructure and the confidence of its investors in Firmus’ execution capability. Nvidia’s involvement, in particular, signals potential synergies in hardware provisioning and co‑marketing of AI workloads, while Jane Street’s stake underscores confidence in the longer‑term economics of AI‑centric data centre assets.

Risks and Execution Challenges

Despite the strong headline numbers, several risk factors remain. First, the concentration of revenue on a single anchor customer creates exposure to OpenAI’s demand‑cycle decisions. If OpenAI adjusts its compute roadmap or shifts to alternative providers, Firmus could see a material dip in utilisation.

Second, the 900 MW figure represents contracted capacity, not actual power generation. The Malaysian sites are still under construction, and the timeline for connecting to the grid could be affected by regulatory, supply‑chain, or weather‑related factors. Until the sites deliver electricity, no revenue is earned, meaning the company must continue to fund construction and operating expenses from cash reserves or external financing.

Third, the pre‑sell model, while de‑risking revenue after power is online, does not eliminate execution risk. Firmus must meet performance standards stipulated in the OpenAI contract, including latency, reliability, and security benchmarks that are higher than the industry average for standard enterprise data centres.

Implications for the Pre‑IPO Thesis

The combination of a high‑profile anchor customer, a large and fully contracted capacity pool, and a backer roster that includes Nvidia and major private‑equity firms creates a compelling narrative for a near‑term public listing. Investors evaluating the pre‑IPO market will likely focus on the timing of power delivery relative to the fiscal year, the ability to convert contracted megawatts into billable revenue, and the durability of the OpenAI relationship.

If Firmus can bring the Malaysian sites online as projected, the company would transition from a growth‑stage, cash‑burning business to one with a predictable, recurring revenue base—a shift that is typically rewarded in the public markets. Conversely, delays or a loss of anchor demand would raise questions about the valuation premise.

Conclusion

Firmus’ multi‑year OpenAI agreement pushes its contracted AI‑compute capacity beyond the 900 MW threshold, a scale that is rare for a private data‑centre operator. The deal, together with a $10.5 billion-plus valuation backed by Nvidia, Jane Street, Blackstone, and Coatue, positions the company as a prominent candidate for a 2026‑2027 IPO. However, the investment case hinges on successful execution of the Malaysian build‑out and the sustained demand from OpenAI. Analysts should monitor grid connection milestones, utilization performance, and any renegotiation of anchor terms as key inflection points.