AI Chips Heat Up as Energy Crunch Looms

Cerebras Systems raised $5.55 billion in its initial public offering Wednesday, pricing shares at $185 each—well above the expected range—as the AI chipmaker capitalized on surging demand for alternatives to Nvidia's dominance , according to CNBC. At the IPO price, Cerebras is now valued at $56.4 billion on a fully diluted basis , making it one of the largest U.S. IPOs ever and giving the company a market valuation of more than $56 billion , Kiplinger reported.

The offering comes as global electricity consumption for data centers is projected to double to reach around 945 TWh by 2030 , according to the International Energy Agency. Lawrence Berkeley National Laboratory's report finds that data centers consumed about 4.4% of total U.S. electricity in 2023 and are expected to consume approximately 6.7 to 12% of total U.S. electricity by 2028, with total data center electricity usage climbing from 58 TWh in 2014 to 176 TWh in 2023 and estimates of an increase between 325 to 580 TWh by 2028 , the Department of Energy announced.

Can the Grid Keep Up With AI's Appetite?

The energy crunch is already reshaping corporate strategy. Cisco shares soared 17% in extended trading Wednesday after the networking company issued results and guidance that topped Wall Street's projections, announcing it's cutting its workforce this quarter by fewer than 4,000 jobs, representing less than 5% of total employees, with revenue increasing 12% in the quarter ended April 25 to $15.84 billion , CNBC reported. Cisco has taken $5.3 billion in AI infrastructure orders from hyperscalers so far this fiscal year, and raised its full-year order expectation to $9 billion from $5 billion previously , according to Reuters.

Data centers are projected to account for 38% of net electricity consumption through 2037, driven by aggressive hyperscaler capital expenditure and the accelerating energy intensity of artificial intelligence workloads , the National Electrical Manufacturers Association said in a forecast published May 7. Data centers accounted for 17% of electricity demand growth worldwide last year, according to the IEA report, compared with around 50% in the U.S. , Fortune reported.

The geographic concentration is stark. According to a Bloomberg News analysis, data centers accounted for almost 40 percent of Virginia's total consumption in 2024 , Consumer Reports found. That same Bloomberg analysis found that areas with high concentrations of data centers saw electricity prices jump 267 percent over the past five years .

Are Efficiency Gains Real or Just Hype?

Researchers are racing to address the energy problem. Tufts University researchers have unveiled a radically more efficient approach that could slash AI energy use by up to 100× while actually improving accuracy , according to ScienceDaily. By combining neural networks with human-like symbolic reasoning, their system helps robots think more logically instead of relying on brute-force trial and error .

Energy efficiency in large language model inference has improved 100,000x in the past 10 years—demonstrating that accelerated computing is sustainable computing , NVIDIA stated. NVIDIA has a long history of driving performance and energy efficiency, with the number of tokens generated within the same power budget increasing by more than 1 million times from the NVIDIA Kepler GPU in 2012 to the NVIDIA Vera Rubin platform this year , the company announced at CERAWeek.

But the efficiency gains may not be enough. AI-optimized servers are fueling the increase in data center power consumption, with their electricity usage set to rise nearly fivefold, from 93 TWh in 2025 to 432 TWh in 2030 , Gartner reported. In 2025, AI-optimized servers are projected to represent 21% of total center power usage and 44% by 2030, representing 64% of the incremental power demand for data centers .

Will AI Solve Its Own Energy Problem?

The KPMG global study of more than 1,200 energy leaders across 20 markets offers their view that AI's potential climate benefits far outweigh its energy footprint , the World Economic Forum reported May 13. The KPMG study indicates that by 2027, 62% of respondents believe major data and AI operators expect to self-generate clean energy, investing directly in renewables .

If a single power plant could increase production by 5% with the help of AI and digitalization tools, scaling those efficiency benefits across 10 power plants would essentially create 50% net-new power generation without placing a single steel beam in the ground, and applying this example at the national level would mean that countries looking to secure their power grids could unlock significant hidden production capacity , according to a World Economic Forum analysis published in January.

Morgan Stanley Research forecasts U.S. data center demand could reach 74 GW by 2028, with a projected shortfall of about 49 GW in available power access, requiring billions in capital for new energy infrastructure . Large technology companies are likely to commit more than $1 trillion of spending in just the 2025-2026 period , Morgan Stanley reported.

Driven by data centre investments, the capital expenditure of five large technology companies surged to more than $400 billion in 2025 and is set to increase by a further 75% in 2026 , the IEA announced. The pipeline of conditional offtake agreements between data centre operators and small modular reactor nuclear projects has grown from 25 gigawatts at the end of 2024 to 45 gigawatts today, and constrained by slow grid connections, data centre developers are also advancing a large number of projects with onsite natural gas-based power generation, largely in the United States .

What Changed This Week

The Cerebras IPO marks a turning point for AI infrastructure investment, with the company's $56 billion valuation reflecting investor confidence that alternatives to Nvidia can capture meaningful market share. Cisco's simultaneous job cuts and AI order surge illustrate how established tech companies are reshaping their workforces to chase hyperscaler spending. Meanwhile, new research on 100x efficiency improvements and grid optimization tools suggests the industry recognizes that energy constraints could become the binding limit on AI growth—and is racing to solve the problem before it becomes existential.

What to Watch

Cerebras shares are expected to begin trading on the Nasdaq Global Select Market on May 14, 2026, under the ticker symbol "CBRS" . The stock's first-day performance will test whether public market investors share the enthusiasm that drove the IPO 20 times oversubscribed. Cisco's latest round of job cuts will begin on May 14 , with the company expecting to recognize about $450 million in restructuring charges in the fiscal fourth quarter. Watch for quarterly earnings from other networking equipment makers to see if Cisco's AI infrastructure order surge is industry-wide or company-specific. The IEA is expected to release updated data center energy projections later this year, which could revise the 2028 consumption estimates that are currently driving utility planning and grid investment decisions.