AI demand pulls in power chips
- AI infrastructure spending is expanding beyond accelerators into power, management, and industrial semiconductors. - GE Vernova raised its 2026 revenue and margin outlook, citing surging data-center demand as a driver. - The shift means system components are now driving orders and complicating hardware forecasting and BOM planning. ( )
Artificial intelligence spending is now lifting the chips that feed and regulate power, not just the processors that do the computing. (theregister.com) Texas Instruments said on April 22 that first-quarter 2026 revenue rose 19% from a year earlier to $4.83 billion, with chief executive Haviv Ilan saying growth was led by industrial and data center demand. TI forecast second-quarter revenue of $5.0 billion to $5.4 billion and earnings of $1.77 to $2.05 a share. (ti.com; channelnewsasia.com) Reuters reported that TI’s outlook beat Wall Street estimates of $4.86 billion in revenue and $1.57 a share in profit for the June quarter. Reuters also said TI shares rose more than 10% in extended trading after the forecast. (money.usnews.com) These chips do the basic electrical work inside a server: they convert current, stabilize voltage, and move signals from sensors into digital data other semiconductors can use. Reuters said that is the category TI sells into data centers as cloud companies keep building out artificial intelligence systems. (wTVBam.com) The same build-out is showing up in power equipment. GE Vernova raised its 2026 revenue forecast on April 22 to $44.5 billion to $45.5 billion from $44 billion to $45 billion, after first-quarter orders jumped more than 71% organically to $18.3 billion. (money.usnews.com) GE Vernova also lifted its 2026 adjusted core profit margin forecast to 12% to 14% from 11% to 13%, helped by demand from data centers for power and electrification equipment. Reuters said the company’s electrification unit posted $528 million in core profit, up from $205 million a year earlier, while power unit profit rose nearly 57% to $811 million. (y94.com) The Register reported on April 23 that the supply squeeze has widened from high-bandwidth memory into power-management chips and baseboard management controllers, the small control chips that monitor and manage servers. That means artificial intelligence hardware is pulling in more of the bill of materials around each rack, not just the headline accelerator. (theregister.com) DigiTimes reported on April 23 that TI’s results pointed to a stronger recovery in industrial demand alongside artificial intelligence demand, giving suppliers another signal that data-center spending is spreading into older analog and embedded categories. (digitimes.com) For hardware makers, that shifts the bottleneck from one famous chip to a larger parts list. The next question is no longer only how many accelerators can ship, but whether the power, control, and grid gear around them can arrive on time too. (theregister.com; money.usnews.com)
