AI Industry Shifts Toward Inference and Advanced Packaging Innovation
NVIDIA CEO Jensen Huang predicts that demand for AI inference computing could grow a billionfold, shifting the industry's focus from costly model training to large-scale inference that supports search, generative AI services and robotics. Silicon photonics is helping overcome data-transfer bottlenecks, while TSMC's CoWoS advanced packaging integrates computing components with high-bandwidth memory, making it critical to maintaining a lead in computing power.
As of July 19, 2026, CompuForum 2026 had identified memory, storage and systems integration as new battlegrounds in the AI race. Adobe's 2026 report said the proliferation of AI was eroding the value of content volume, making human taste and judgment more valuable. The reports provided no figures for CoWoS investment, capacity-expansion dates or Adobe's latest financial results, preventing further quantification.
All Coverage
5 original reportsThe Backstory
The history behind this eventAI Reshapes Global Industries, From Executive Turnover to Advanced Chip Packaging
Generative AI is moving from cloud-based model training to on-device inference, prompting companies to overhaul their decision-making and operations teams while upgrading computing infrastructure. Leadership changes at Coca-Cola and Walmart, alongside Nvidia's and TSMC's investments in silicon photonics and CoWoS, show that AI is reshaping corporate management, chip interconnects and competition in advanced packaging at the same time.
Recent reports show AI applications shifting to local devices such as desktop computers to reduce the risk of data leaving the device and shorten response times. The corporate sector is also seeing a wave of leadership transitions, with the presidents of Coca-Cola and Walmart stepping down in succession. As inference demand expands, Nvidia and TSMC continue to develop high-speed data transmission and CoWoS technologies. Available information does not disclose specific dates or investment amounts.
Subscribe to Mark Radar Weekly
Every Friday, the week's strongest signals in your inbox. Unsubscribe anytime.