The balance of power in AI data center lease negotiations is undergoing a dramatic reversal. Historically, hyperscale cloud providers like Microsoft and Google dominated discussions with emerging cloud service firms such as CoreWeave, Nebius, and Nscale, leveraging their immense capital reserves and top-tier credit ratings. However, as major cloud players race to deploy Nvidia server racks, data center operators are gaining significant leverage, and the terms of contracts are now tilting in favor of the supply side.
This transformation is reshaping the profit dynamics across the entire AI infrastructure chain—from the strictness of service level agreements to electricity pricing, and even the role of chip manufacturers. These changes are rewriting the foundational rules of the industry.
Data Center Operators Gain Leverage, Payment Terms Tighten
Service level agreements are emerging as the primary battleground in these negotiations. Previously, large cloud providers opened talks by demanding near-perfect uptime for every server rack and imposing stringent controls on facility temperature and humidity. A data center executive revealed he had encountered contracts where a single rack outage—whether from a power failure, overheating, or switch malfunction—would allow the cloud client to cancel six months of rent. Moreover, if SLA breaches accumulated past a certain threshold, the cloud firm could unilaterally terminate the entire lease.
An industry executive noted that negotiating SLAs is fundamentally a trade-off between optimal pricing and contract durability—the stricter the terms, the higher the price, but also the greater the risk. Securing an SLA with lighter penalties can be worth accepting, even if it means settling for lower rates. Currently, with operators holding more cards, these extreme provisions are gradually being softened.
Payment terms are also shifting in the operators' favor. A credit executive described witnessing a case where a client leased only a small portion of a massive data center, yet the contract stipulated that a failure to pay on time would obligate the client to cover the entire building's rental costs for a fixed period. The data center owner themselves acknowledged the aggressiveness of the demand, admitting, "Look, we know this is outrageous... but we can get away with it."
Chip Makers Enter the Fray as Credit Guarantors
Another factor boosting operators' leverage is the proactive involvement of chip manufacturers. According to another data center executive quoted in the report, Nvidia and AMD are sometimes competing for the same data center project, vying to offer credit guarantees to clients to ensure that facilities housing their chips are successfully built. Nvidia is proving more aggressive because of its stronger balance sheet, noted the source. These arrangements are particularly advantageous for developers, as Nvidia and AMD are willing to back leases with credit support for up to 15 years, compared to the six-year terms Nvidia has previously disclosed with some emerging cloud providers.
Electricity Costs: Up to 400% Variance in a Single Month
Electricity represents the largest individual operating cost for data centers, sometimes exceeding 20% of total expenses, though it can occasionally be far higher. The credit executive observed that in the same month, electricity prices could vary by as much as 400% across different customers. The paradox, he pointed out, is that every client believes they have secured a favorable rate, yet in reality, the largest cloud firms often pay the highest prices—simply because they have the financial capacity to do so. The pressure of power shortages is also spreading, with reports indicating Elon Musk anticipates a significant electricity shortfall by 2027 and is planning to build his own turbine blade factory. Meanwhile, some developers are paying hefty premiums for gas turbines from dubious "fly-by-night" companies, further complicating insurance and financing costs.
These mounting factors are making data center cost forecasting extremely difficult. Over the past few years, the cost of building a one-gigawatt data center has more than doubled, and if prices continue to climb, investor sentiment across many global public companies could face considerable strain.
Compute Fragmentation: Nvidia's Strategic Play
Amid the tight supply of large-scale data centers, Nvidia CEO Jensen Huang outlined a different approach at the Equinix customer conference. Compute is fragmenting, he said, and the AI world will fundamentally decentralize and become highly distributed. At the event, Nvidia, Equinix, and Together AI announced a joint initiative in which Together AI will purchase Nvidia hardware, deploy it in Equinix's existing facilities, and offer inference services for open-source AI models to small and medium-sized businesses. Huang described the logic as operating AI tasks with low latency tolerance in the Equinix data centers nearest to users, while memory and inference tasks can be handled from more remote locations.
Executives from Google, Cisco, and the emerging cloud provider Lambda Labs echoed Huang's views, suggesting that AI inference will increasingly become decentralized, and a distributed network of thousands of smaller facilities could emerge as the dominant model going forward.
Comments