Oracle’s expanded Bloom Energy agreement for up to 2.8GW of fuel cells underscores the pressure to secure time-to-power for AI clusters, while operators weigh outage exposure, capital discipline and tightening energy reporting rules.
CETA System Co., Limited is drawing attention to Oracle’s expanded fuel-cell procurement with Bloom Energy, lifting the framework to 2.8GW as operators seek distributed power options for high-density AI data centres.
An initial 1.2GW tranche is being deployed, and Bloom has previously delivered a fully operational system to Oracle in 55 days, ahead of a 90-day expectation. Lee Tsz-Hin (李子軒), Chief Executive Officer of CETA System Co., Limited, calls it “a procurement-led response to grid lead times that can no longer set the pace for AI capacity”.
Bloom Energy disclosed earlier in the current reporting window that deployment of the first 1.2GW of contracted capacity has commenced across Oracle projects in the United States, with installation activity scheduled to continue through the following year.
Oracle positions the buildout as supporting cloud-infrastructure demand from customers running AI and other compute-intensive workloads; Mahesh Thiagarajan, Executive Vice President at Oracle Cloud Infrastructure, referred to “rapidly deploying Bloom’s reliable, efficient fuel cell energy” to meet those requirements.
Following the disclosure, Bloom Energy shares rose 12.6% to $221.8 in extended trading, while Oracle shares were reported up 7%.
CETA System also points to a widening power and cooling gap as GPU-driven rack densities rise. Average rack density has moved from 8.4 kW at the start of the decade towards projected 30 kW configurations within the next few years, while some AI workloads are pushing beyond 100 kW per rack.
Traditional air cooling reaches practical limits at about 20 to 25 kW, and with cooling accounting for around 40% of data-centre energy use, operating cost and resilience become tightly linked. “The commercial risk shows up first as thermal constraint and then as downtime,” Lee notes.
In the latest global estimate, data centres accounted for around 1.5% of electricity consumption, equivalent to 415 terawatt-hours, with projections pointing to more than 945 TWh by 2030. In the most recent United States tally on record, data-centre electricity use reached roughly 176 TWh, about 4.4% of national demand, with projections indicating growth to between 325 and 580 TWh later in the decade.
Outage economics sharpen the focus on reliability. Survey data compiled by the Uptime Institute indicates most operators report their most recent significant outage costing more than $100,000, with roughly one in five incidents exceeding $1 million.
Benchmarks from Cushman & Wakefield estimate an average cost of roughly $11.73 million to develop one megawatt of data-centre critical load capacity in the United States, a level of capital intensity that makes electrical failures disproportionately expensive to recover from. Power-related failures account for more than one-third of outages.
Operational intelligence is increasingly treated as a resilience and compliance tool rather than a feature. Real-time monitoring across electrical, mechanical and environmental systems can detect anomalies before escalation, while predictive maintenance has been associated with 30% to 50% downtime-risk reduction and 20% to 40% longer equipment life across industrial applications.
Advisory-first deployments configure AI models to facility topology and infrastructure parameters and produce recommendations from sensor data, while operators retain supervision and approval. “Advisory-first only works when operators keep the final call,” Lee says.
On-site generation strategies increasingly connect those operational priorities with procurement. Solid oxide fuel cell systems are presented as fuel-agnostic, accommodating hydrogen, natural gas and biogas, as well as ammonia, methanol and syngas, and modular designs can scale from around 100 kW units to multi-megawatt clusters.
Fuel-cell installations are described as delivering power three to seven years ahead of traditional utility connections, deploying four to eight times faster than alternative on-site generation technologies, with certain configurations reaching full operation within six months.
Efficiency claims cited for this class of system include 60% to 65% typical electrical output, up to 75% stack-level electrical efficiency in advanced designs, and a 10% to 15% power usage effectiveness improvement through absorption-chiller heat recovery into liquid-cooling loops; PUE, power usage effectiveness, measures total-to-IT energy.
Regulation is tightening the link between engineering choices and disclosure. The revised EU Energy Efficiency Directive sets monitoring and reporting obligations for data centres with installed IT power demand of at least 500 kW, requiring disclosure of energy performance indicators to a European database.
Germany’s Energy Efficiency Act is set to require 100% renewable electricity coverage from the start of 2027, a maximum annual PUE of 1.2 for facilities commencing operation from mid-year, and energy reuse factors reaching 20% by mid-2028. The Corporate Sustainability Reporting Directive, the EU Taxonomy and the Climate Neutral Data Centre Pact further reinforce sustainability reporting expectations, including a climate neutrality target by 2030.
For investors and operators, the combined pressure of time-to-power, outage exposure and reporting obligations is sharpening the focus on measurable performance and disciplined procurement. CETA System expects distributed generation and operator-supervised optimisation to remain central as AI compute density continues to rise.
About CETA System
CETA System Co., Limited is a Hong Kong-incorporated technology company founded in 2017. Its vendor-agnostic platform delivers AI for data-centre infrastructure, combining HVAC and chiller-plant energy optimisation with predictive maintenance for UPS systems, generators and chillers, integrating with building-management and DCIM environments in an advisory-first deployment model across Asia-Pacific and beyond.
• Website: https://cetasystem.com
• Registered business: CETA System Co., Limited (Hong Kong BRN 67731517; CRN 2533166)
