Who will succeed in making Nvidia falter the most, the new tech challenge since mid-2026.

Should I buy the dip on Monday? Take in account the Paradigm Shift of June 2026. On April 6, Broadcom filed an 8-K with the SEC—the kind of routine regulatory document most people skip past. But this one contained something remarkable: a new five-year agreement with Google to develop and supply custom AI chips for future generations of its Tensor Processing Units (TPUs) — custom chips optimized for AI workloads — through 2031. ByteDance announces it is designing custom processors for AI inference to reduce its reliance on third-party hardware in the face of rising costs and chip shortages. In partnership with Samsung and TSMC, the company plans to begin production in 2026.

And also: A massive consortium called UALink (Ultra Accelerator Link) has emerged, driven by AMD, Broadcom, Intel, Meta and Microsoft. The goal is to create an open connection standard between AI accelerators. In April 2026, the consortium released version 2.0 of the UALink specification. This allows server manufacturers, for the first time, to use non-Nvidia switching chips to connect up to 1,024 accelerators within a single pod.