Power Crisis : AI Energy Boom
Chart of the Week:
Nvidia Delivers Strong Results but Tech Faces Valuation Scrutiny
Nvidia briefly reignited market optimism as 3Q revenue reached USD 57bn, up 22% QoQ and 62% YoY, beating expectations on improved supply-chain throughput and stronger-than-expected GB300 sales versus GB200. Data-center revenue hit a record USD 51.2bn—about 90% of total—up 25% QoQ. GPU-related revenue rose 27% QoQ, dispelling concerns about slowing GPU demand, while networking revenue surged 162% YoY. Both segments set new historical highs. Better supply-chain efficiency also narrowed the gap between orders and deliveries, supporting earnings upgrades. Nvidia’s robust guidance triggered a sharp AI-stock rebound. However, accounts receivable rose sharply—from USD 23.065bn to USD 33.391bn—reviving concerns about whether AI workloads can generate enough profit to justify massive capex. AI-bubble worries persist.
Market Recap 1:
Financial Assets Pull Back After Strong Gains; Market Short-Term Focus on Liquidity and Rising Volatility
Initial jobless claims in mid-October reached 232,000, the highest in two months. September nonfarm payrolls increased by 119,000—better than expected—while the unemployment rate rose to 4.4%, the highest since October 2021. Weak economic data, combined with declining expectations for a December rate cut and lingering doubts about AI monetization, triggered a pullback in financial markets. The Philadelphia Semiconductor Index (SOX), with the richest valuations, saw the sharpest correction as investors shifted attention to liquidity conditions and rising volatility.In Japan, the government is preparing a comprehensive economic package—including a ¥20,000 child allowance—with a total scale estimated above ¥20 trillion. The large-scale fiscal stimulus raised concerns over inflation and fiscal deterioration, pushing Japan’s 10-year government bond yield to 1.8%, the highest in roughly 17.5 years. The yen weakened, and Japanese equities corrected alongside global markets.
Market Recap 2:
Federal Reserve Officials Divided; U.S. Dollar Returns to 100 Level, Commodities Under Pressure
Federal Reserve Governor Waller voiced support for a December rate cut, while Vice Chair Jefferson stressed the need for caution — highlighting growing divergence within the Fed. Meeting minutes also revealed that some officials warned a rate cut could intensify inflation pressures on the economy, and views on inflation and labor-market conditions remain split. Additionally, due to the prior U.S. government shutdown, the October nonfarm payroll report will now be released together with the November data after the December FOMC meeting. With limited visibility on economic conditions, markets sharply reduced expectations for a December rate cut. Fed funds futures now imply only a 39.1% probability of a 25 bps cut, while 60.5% expect no action. U.S. Treasury yields moved higher across the curve.
What’s Trending:
Fed December Rate-Cut Expectations Decline; Gold Faces Short-Term Pressure but Long-Term Outlook Intact
Several previously dovish-leaning Federal Reserve officials turned ambiguous or even hawkish this week. Combined with Wednesday’s release of the October meeting minutes — which showed that “most” officials prefer keeping rates unchanged for the rest of the year — market expectations for a December rate cut have dropped to below 30%, compared with nearly 100% just a month ago. Even so, expectations for rate cuts in 2026–2027 have remained largely steady. The 2-year Treasury yield hovered around 3.60%, while the 10-year yield stayed near 4.14%, leaving the yield curve broadly stable.
In Focus 1:
The Race for Computing Power Has Become a Race for Electricity; Demand for AI Infrastructure Surges
Traditional communication infrastructure focused on base stations and fiber networks to transmit information. Today, data centers and AI chips—driven by massive computing requirements—have emerged as the new core infrastructure. According to the U.S. Department of Energy, data centers currently consume about 5% of total U.S. electricity, a figure projected to rise to 12% by 2030, highlighting a major structural shift in energy demand. Global AI-related power consumption is estimated at 44 GW in 2025, rising to 156 GW by 2030, an increase of more than 3.5 times. The imbalance between power supply and demand primarily stems from slow growth in power generation capacity—limited by plant construction speed, efficiency constraints, and clean-energy requirements—far below the explosive increase in AI-driven electricity usage.
In Focus 2:
AI Application Output Set to Multiply, Driving Earnings Growth in Utilities
AI-related applications are expected to see exponential growth in market value, requiring large-scale and sustainable power to operate chips and data centers. Many AI data centers demand power at the megawatt level, meaning the bottleneck has shifted from chip shortages to power shortages. Adequate grid capacity and interconnection permits have become critical considerations for new data center development. As a result, major tech companies are signing long-term power purchase agreements (PPAs) with large utilities, supporting stable earnings and positioning utility companies for continued profit growth next year. Beyond computing inside AI server racks, the transmission of processed data to large language models, software platforms, and industry applications is driving demand for external infrastructure such as optical communications and networking. This forms an interconnected AI ecosystem that also includes emergency backup systems, integrated energy storage, and rack-level battery backup units (BBUs). Battery energy storage systems (BESS) are also expected to expand at a compound annual growth rate of around 20%.
