Who is financing AI's power backbone?

The answer: private capital is backing the physical bottlenecks

General Catalyst, Koch Equity Development and unnamed co-investors have agreed to provide $2 billion of convertible preferred equity to Axiom Solutions International, the cloud-and-power infrastructure business that Flex plans to separate. Flex gives Axiom an initial enterprise value of $37.5 billion and says the separation is expected in the first quarter of 2027. This is financing for the equipment around AI computing—power conversion, thermal management and scalable infrastructure—not an investment in a new foundation model.

That distinction matters. Model developers attract most public attention, but new data centres also need grid connections, transformers, power electronics, backup systems, cooling and manufacturing capacity. A large preferred-equity commitment signals that investors expect these constraints to remain commercially valuable. It does not show how much of Axiom's future revenue will come from AI workloads, how quickly customers will build capacity or whether projected margins will survive competition and construction delays.[1][2]

The terms show capital is available, but not cheaply

Before the planned separation, the preferred stock carries a 10% annual cash dividend. After separation, the rate steps down to 6% if paid in cash or 7% if Axiom pays in kind by adding the dividend to the security rather than distributing cash immediately. The terms are subject to increases after the fifth anniversary and other adjustments. On a $2 billion principal amount, a 10% rate implies $200 million a year before considering timing, adjustments or any conversion mechanics not detailed in the announcement.

Preferred equity sits ahead of ordinary shareholders for its contractual economics and can reduce near-term dilution compared with issuing common shares, but its dividend burden is real. Paying in kind can conserve cash after separation while increasing the amount owed. The announced enterprise value should not be mistaken for a public-market price discovered through trading. It is the transaction's stated starting valuation and remains dependent on the financing and separation completing on the disclosed terms.

General Catalyst would also receive the right to nominate one director to Axiom's board after the separation. That governance right gives the lead investor influence alongside the economic protections. Flex did not disclose the complete investor allocation, conversion price, liquidation preference or all adjustment provisions in the press release and exhibit, so readers cannot reconstruct the full return profile from these documents alone.[1][2]

Where the money is intended to go

Flex says net proceeds may fund part of Axiom's pending acquisition of EPC Power, repay bridge or intermediate financing used for that purchase, pay dividends on the preferred stock or support general corporate purposes. Flex says it has separately arranged committed term-loan financing for the remaining acquisition cost. In other words, the capital is partly acquisition finance and balance-sheet preparation for the spin-off, not a ring-fenced promise to build a specified number of data-centre sites or add a disclosed quantity of power capacity.

That is why the announcement should not be translated into a claim that $2 billion will become $2 billion of new AI infrastructure. Money can refinance an earlier obligation, pay transaction costs or support working capital. The documents provide no denominator for additional megawatts, cooling units, factory capacity, customers or deployment dates attributable to this investment. Those operating measures will matter more to communities, grid planners and customers than the headline valuation alone.[1][2]

What changes for workers, customers and communities

If Axiom expands, manufacturers and suppliers may see demand for electrical equipment, cooling systems, assembly and field services. Data-centre customers could gain another scaled supplier able to combine power, thermal and compute infrastructure. Yet a corporate separation can also create uncertainty over reporting lines, benefits, systems and capital allocation. Flex itself lists employee departures, management distraction, integration problems and disruption to customer and supplier relationships among the risks.

For communities, the financing is several steps removed from a local project. A better-capitalised supplier may enable faster construction, but each facility still needs land, grid capacity, water or alternative cooling, permits and a credible plan for who pays for network upgrades. No community should treat a financing announcement as evidence that a specific project is approved, low-carbon or beneficial. Those conclusions require site-level capacity, energy, water, jobs and cost-allocation evidence.[1]

What the filing proves—and what it leaves open

The issuer announcement and SEC exhibit establish that an agreement was announced on 5 October, identify the investors named by Flex, state the principal amount and initial enterprise value, and disclose headline dividend and governance terms. They also state that closing requires regulatory approvals and customary conditions. They do not prove that the investment has closed, that the EPC Power acquisition will complete, that the separation will occur on schedule or that either company will achieve the expected strategic and financial benefits.

Both documents are company-originated. Filing an announcement with the SEC makes it an official public record, but it does not turn management's projections into independently verified outcomes. The risk section explicitly warns that timing, approvals, integration, ratings, tax treatment, capital structure, employee retention and market conditions could change the result. It also directs investors to the Form 10 and future proxy materials for fuller information.[1][2]

What would change the assessment

The assessment would strengthen when the investment and EPC Power acquisition close, final terms are filed and Axiom publishes audited financial statements that separate AI-related infrastructure demand from the rest of its business. Useful operating evidence would include customer concentration, backlog quality, capacity delivered on time, power and cooling efficiency, warranty performance, capital expenditure, cash generation and the share of revenue tied to completed rather than proposed data-centre projects.

It would weaken if approvals are delayed, the separation slips, preferred dividends consume more cash than expected, bridge financing grows, major customers defer construction or integration costs erode the anticipated benefits. Until closing and audited performance data arrive, the defensible conclusion is narrow: sophisticated investors have committed substantial capital on protected terms to a company positioned around AI's physical infrastructure bottlenecks. The return and public benefit remain unproven.[1][2]