AI Data Centers Are Maxing Out P&C Limits, AIG's CEO Says

Eric Andersen, ten weeks into running AIG, told Bloomberg TV this week that the AI data center buildout "is absolutely maxing out the P/C insurance industry in terms of the limits that are required" (Insurance Journal, August 2026).

S&P Global Ratings puts numbers behind the claim: roughly $10 billion of new insurance premium in 2026 alone, about twice the entire global aviation insurance market (S&P via Beinsure).

Key Takeaways

$20 billion to $30 billion of total insurable value per hyperscale campus, with the construction phase alone running $10 billion to $30 billion. Those are values the property market last confronted in offshore energy.
Roughly 30 distinct insurance products span one project's life, from permitting and financing through construction into operations, which is the basis of AIG's claim that few carriers can serve a whole program.
Property catastrophe rates fell 16% at the mid-year renewals. Data center capacity is the conspicuous exception in a market softening nearly everywhere else.
$3.5 billion is where Aon's Data Center Lifecycle Insurance Program stood in April after adding a billion of capacity in one step, against billions of dollars of construction risk still running uninsured.

The Exposure, Sized

S&P counts roughly 11,000 data centers in operation globally, an insurable asset base above $2 trillion, with annual investment in new construction on track to pass $300 billion by 2027. The number that binds capacity is the per-site one: a single hyperscale campus carries estimated total insurable values of $20 billion to $30 billion, with the construction phase alone at $10 billion to $30 billion (Artemis, 2026).

Those are values the property market last met in offshore energy and mega-infrastructure, and it met them with syndication structures built over decades. The buildout is asking for comparable limits on dozens of concurrent projects, on timelines compressed by hyperscaler urgency, for occupancies whose loss behavior has a few years of history at current scale.

Andersen's inventory of the coverage stack is worth taking literally. On AIG's second-quarter earnings call he described roughly 30 distinct insurance products a project needs across its life: builder's risk, marine cargo for equipment in transit, general and excess liability, cyber, business interruption and power-generation covers among them (AIG Q2 2026 earnings call). AIG writes across that stack and insures the power providers feeding the sites, which is the basis of his claim that few carriers can serve the whole program.

What "Maxing Out the Limits" Means Mechanically

Large individual carriers typically deploy line sizes in the hundreds of millions on a single complex property risk, and even the most aggressive rarely put more than a billion of net limit on one location. Building a $25 billion tower at those line sizes requires stacking dozens of participations across quota shares and excess layers, with every carrier in the tower watching its aggregate across every other data center program it has already joined.

The refinery and offshore-energy schedules that ran this playbook arrived a handful of projects at a time. A $300 billion annual construction pace asks the same syndication machinery to assemble multiple such towers per quarter, drawing on the same finite carrier capacity, the same reinsurers and increasingly the same facultative markets. The last billion of a mega-tower is where scarcity shows first: hardest to place, least premium per unit of limit, and the first layer a constrained market declines.

The pricing consequence is that experience rating is unavailable at exactly the point where the limits are largest. Nobody holds credible frequency and severity distributions for a $25 billion campus of GPU halls. Legacy server-farm fire experience transfers imperfectly to liquid-cooled, high-density clusters, the business-interruption profile depends on chip supply chains with single-vendor concentration, and the liability exposure of facilities training frontier models has no case law. Rate has to come from values, construction characteristics, redundancy engineering and named-peril accumulation, carrying an uncertainty load that has to stay defensible three years out.

That is harder to hold to because the segment is the only hard market left. Property catastrophe rates fell 16% at the mid-year renewals and casualty programs are fighting for rate against social inflation. Scarcity pricing and a genuine uncertainty load both produce high premiums; only one produces adequate ones. Pricing without loss history is what the profession did for cyber in 2015 and gene therapy in 2023.

Accumulation is the same problem seen from the portfolio side. One campus stacks construction all-risk, property and equipment breakdown, business interruption, cyber with physical consequences, and liability following whatever the tenant trains, all on one footprint. A single convective storm, fire or grid failure can trigger several coverages inside the same tower of limits.

The Buyers Are Bigger Than the Sellers

The uninsured remainder is both the confirming evidence and the constraint. ENR reports billions of dollars of data center construction risk currently proceeding without full coverage, which in a rational market means the price of the missing layers exceeded what sponsors would pay, or the capacity did not exist at any price. Hyperscalers with trillion-dollar balance sheets can self-insure gaps that would stop a conventional developer cold.

That inverts the usual bargaining structure. The buyers are larger and better capitalized than the sellers, which caps how much rate scarcity can extract even while demand outruns supply. S&P expects the operational-phase slice with the least modelable tail, business interruption and technology losses, to stay substantially self-insured or migrate to captives, and expects insurance-linked securities to take a growing share (Artemis, 2026).

That is the market conceding it cannot yet price the layer where actuarial method has least to stand on. Standard BI valuation starts from historical revenue and a recovery period. A facility whose tenant's revenue did not exist three years ago, and whose critical equipment has a supply queue rather than a spot market, defeats both inputs. A carrier writing the layer anyway is exposed on its recovery-period assumption before its frequency assumption.

The plumbing is being built around that gap rather than through it. Aon expanded its Data Center Lifecycle Insurance Program to $3.5 billion in April, adding a billion of capacity in one step (ENR, 2026). Andersen spent his career until June at Aon, most recently as its president, so the executive now describing carrier capacity limits helped design the broker structures aggregating demand against them.

AI Data Centers Are Maxing Out P&C Limits, AIG's CEO Says

Key Takeaways

The Exposure, Sized

What "Maxing Out the Limits" Means Mechanically

The Buyers Are Bigger Than the Sellers

Further Reading

Sources