AI Needs Terawatts of Power, Not Just Next-Gen Chips

Rendy Andriyanto
Rendy Andriyanto
Gotrade Team
Reviewed by Gotrade Internal Analyst
AI Needs Terawatts of Power, Not Just Next-Gen Chips

Share this article

Gotrade News - Power demand for AI data centers became the central narrative across Q1 2026 prints in energy and infrastructure. Energy Transfer reported record revenue of $27.77 billion and argued AI needs terawatts of energy, not just next-generation chips.

The shift extends the AI investment theme from chip names into power and midstream operators. Investors are now mapping the full energy supply chain that has to keep pace with hyperscaler spending.

Key Takeaways

  • Hyperscaler 2026 capex is tracking near $1 trillion annually by 2027, with $567B in new RPO added in the last three months.
  • Energy Transfer reported record $27.77B Q1 revenue with multi-year natural gas contracts to Oracle data centers.
  • Celestica launched 1.6TbE switches for GenAI/ML infrastructure while Cloudflare partnered with Wiz on AI security.

Hyperscaler Capex Approaches $1 Trillion

Hyperscaler annual capex is projected to approach $1 trillion by 2027 according to recent research. The Big Four (Alphabet, Microsoft (MSFT), Amazon (AMZN), Meta) added $567 billion to RPO in just the last three months.

Specific RPO snapshots show Microsoft at $625 billion, Alphabet (GOOGL) at $460 billion, and Amazon at $364 billion. New commitments average roughly $200 billion per month combined across the group.

Token production was up 60% quarter-over-quarter, but related revenue grew only 13%. That divergence indicates token prices are deflating as competition intensifies across model providers.

Energy Transfer Becomes Indirect Beneficiary

Energy Transfer reported record Q1 revenue of $27.77 billion with adjusted cash flow growth of 16.9%. Net margin compressed slightly to 4.9% on cost and raw material volatility.

The company has secured multi-year natural gas contracts to AI data centers including with Oracle. That positioning makes ET an indirect beneficiary of hyperscaler capex without taking direct chip-cycle risk.

The main risks remain Middle East geopolitical tensions and inflation pressuring midstream capex costs. The base case still points to sustained AI-driven energy demand through the decade.

Other Supporting Names on the Tape

Celestica (CLS) announced availability of 1.6TbE switches for GenAI and ML infrastructure. High bandwidth networking is core to modern training cluster topologies.

Cloudflare (NET) partnered with Wiz to secure AI applications against shadow AI. The deal extends the cybersecurity theme into AI governance for enterprise deployments.

What To Watch

Watch whether AI capex ROI begins showing up in cloud revenue prints over the next few quarters. Research notes that free cash flow for five of the seven Magnificent Seven names could turn negative in 2027.

Mid-cycle investors should weigh power and midstream exposure as a complement to chip exposure. As capex rises, beneficiaries extend beyond NVDA and AVGO into power supply chains and networking gear.

Sources

Disclaimer

Gotrade is the trading name of Gotrade Securities Inc., which is registered with and supervised by the Labuan Financial Services Authority (LFSA). This content is for educational purposes only and does not constitute financial advice. Always do your own research (DYOR) before investing.


Related Articles

AppLogo

Gotrade