Nvidia's Spectrum-X Photonics platform is the first CPO Ethernet switch in mass production, delivering 5x lower network power and 64x better signal integrity. Nvidia's networking revenue surged 199% to $15 billion as CoreWeave, Lambda, and Oracle adopt the new platform, cementing a rapidly widening competitive moat. TSM and Lumentum supply the CPO buildout; Lumentum's revenue doubled to $1 billion with multi-hundred-million-dollar orders already locked in for 2027.
It sounds nuts, but SoFi1 is giving new Active Invest users up to $3,000 in stock for a limited time, and all it takes is a $50 deposit to get started.2 See for yourself (Sponsor) Every AI story eventually runs into the same wall: power. Data centers can only pull so much electricity from the grid, and chips can only move data so fast before the network becomes the bottleneck instead of the processor. Nvidia (NASDAQ:NVDA) posted $81.7 billion in revenue for its fiscal first quarter, up 85% year-over-year, with data center revenue alone hitting $75.2 billion.
That kind of growth doesn't come from selling faster chips alone -- it comes from controlling the whole system those chips live in. Now, Nvidia just took its boldest step yet toward owning the wiring, not just the brains, of the AI factory. Nvidia just announced its Spectrum-X Ethernet Photonics platform entered full mass production, becoming the first co-packaged optics (CPO) Ethernet switch built for 200G-per-lane volume shipping.
Traditional AI data centers link GPUs using pluggable optical transceivers -- separate components that convert electrical signals to light and back again. At the scale of a million-GPU cluster, that's a liability: more lasers to fail, more power draw, more heat. SoFi Active Invest is offering a limited-time promotion.
Open an account, fund it with $50 or more, and you could receive up to $3,000 in complimentary stock for Active Invest accounts. See for yourself by clicking here now. Nvidia's fix was to solder the optics directly onto the switch chip.
The results are 4 times fewer lasers, 5x lower network power consumption, 10x better resiliency at scale, and a 64x improvement in signal integrity. Optical loss dropped from roughly 22 decibels to about 4. Translation for shareholders: fewer parts breaking down, lower electricity bills for customers, and a network that keeps pace as clusters scale into the millions of GPUs.
Extract — continue reading at the source.
