New Vulkan-based AI inferencing tests pit Intel’s latest Battlemage architecture against the competition. The results show a landscape in flux, with driver maturity playing a huge role. It’s a fascinating look at the new three-horse race in consumer GPU compute.
The market for packaging data center AI chips is projected to grow at over 40% annually through 2030. While GPU packaging will dominate revenue, the massive investment required poses a major risk to this forecast.
Intel’s VTune Profiler update for version 2025.07 now lists support for the “BMG-G31” GPU. This is the long-rumored, larger Battlemage chip that was thought to be canceled. Its appearance in public patch notes is the strongest evidence yet it’s coming to market.
Analysts say Apple is close to a deal with Intel to have them manufacture a portion of its chips. This would start with low-end Mac/iPad chips in mid-2027 and some iPhone chips by 2028, using Intel’s advanced 18A process.
Early Geekbench results for the unannounced AMD Ryzen 7 9850X3D have surfaced. The chip shows a clock speed boost but its multi-core score is surprisingly lower than the current model. This highlights the pitfalls of taking pre-release benchmarks at face value.
A wild report from Korea suggests Samsung’s own semiconductor division is refusing to supply its consumer electronics arm with RAM unless it pays sky-high prices. Samsung has flatly denied the internal dispute, but the underlying market chaos is undeniable. AI companies are vacuuming up memory, send
Helium isn’t just for balloons. Its unique properties make it essential for cooling, cleaning, and precision in semiconductor manufacturing. With global supply under pressure, the entire tech world has a problem.
In a major shift, Micron is shutting down its entire Crucial consumer division. The company will stop shipping Crucial-branded SSDs and RAM to retailers by the end of its fiscal Q2 in February 2026. This move underscores the overwhelming pull of AI on memory supply chains.
A new startup called Vinci has come out of stealth with $46M in funding and a bold claim: its AI-powered simulation software runs up to 1000x faster than conventional tools. It’s already being tested by major semiconductor companies.
Marvell Technology is looking to make a massive acquisition to supercharge its AI capabilities. The potential $5+ billion deal for Celestial AI highlights the fierce competition to solve a critical bottleneck.
