Machine vision company Cognex buys Intel robotics spinoff RealSense for $600M

NVIDIA (NVDA) has begun production of its Vera Rubin platform, aiming to extend its AI market lead. The platform promises 10x the throughput of its predecessor and is expected to contribute 20% of Data Center revenues in Q3. NVIDIA reported Q2 FY2027 revenues of $96.2B, up 106% YoY, with Data Center revenues at $89B, up 117%. Rivals AMD and Intel are also advancing their AI chip platforms, but NVIDIA's integrated systems remain a key advantage.

Analysts at Susquehanna and Bernstein provided updates on Intel (INTC) and Lumentum (LITE). Christopher Rolland reiterated a Hold rating on INTC with a $115.00 target, while Daniel Zhu maintained a Buy on LITE with a $1220.00 target. INTC has a Hold consensus and $115.23 average target, while LITE has a Strong Buy consensus and $1131.78 average target.

Shares of AI chipmakers fell in Asia and U.S. premarket trading on Tuesday. SK Hynix and Samsung dropped 10.1% and 7.4%, respectively, while U.S. peers like Nvidia and AMD also declined. Broadcom's CEO warned about unrealistic AI investment expectations. Chinese chipmaker CXMT announced advancements in RAM chip technology, potentially challenging major rivals.

Samsung, SK hynix, and other chipmakers join Applied Materials' EPIC Center, a $5B R&D hub in California. The facility aims to accelerate next-gen semiconductor tech development through collaboration. Founding partners include TSMC, Intel, and Micron, with universities and other firms as innovation partners.

Samsung Electronics, SK hynix, and other semiconductor firms joined Applied Materials' $5bn EPIC Center in California for collaborative R&D. The facility aims to speed up next-gen semiconductor tech development, with partners expecting accelerated innovation and commercialization.

Intel officially launched its Crescent Island GPU, now called the Intel GPU D-Series, at the Open Compute Project Global Summit on October 12, 2026. The chip offers 1.3 PFLOPS of FP8 compute, 1.5 TB/s of memory bandwidth, and up to 480 GB of LPDDR5X memory, targeting AI inference workloads. Intel claims a single card can handle 300 billion parameters, with an eight-GPU system supporting over 1 trillion parameters.