The most concrete shift today is ZML’s free LLMD stack, which advertises up to double‑speed inference on today’s leading GPUs, AMD GPUs, Intel Gaudi and even custom ASICs. If the claimed 2× gain holds, operators could halve the number of nodes needed for a given throughput, directly lowering capex and power budgets. The tool is open‑source and drops into existing pipelines, so the cost impact is immediate for any rig in our catalog.
On the hardware‑finance side, SambaNova closed a $1 billion Series F round at an $11 billion valuation. The cash infusion is earmarked for its next‑gen data‑center chip, which could add a new competitor to the Blackwell‑dominant market. While no new SambaNova rigs have appeared in the catalog, the funding signals a potential shift in supply dynamics later this year.
Our rig inventory stayed static: 51 rigs verified, none added in the last 30 days. The lack of new hardware announcements means operators should focus on software levers—like ZML’s optimizer—to squeeze performance from existing stock.
Meanwhile, Anthropic’s hype about Claude achieving “consciousness” offers no technical substance and should be treated as marketing fluff rather than a capability upgrade.
Operators should ask: will the promised 2× speedup translate into real‑world cost cuts on current Blackwell or MI300X servers, or will integration overhead erode the gains?
Composed by the MadCoolStuff editor pipeline · Groq · openai/gpt-oss-120b · 2026-07-08