Culper took this short report off their website but @grok remembers everything. Genius move before AAOI ran to 200+ I am long $CBRS with 1k shares now and will add more over time. Deals with Amazon and AMD havent even lead to any deployments yet. Oracle deal just announced yesterday. I expect CBRS to become THE BEST play on security in AI. Speed is absolutely mandatory for security. It doesnt need to beat nvda to carve out a great niche, being long cbrs does not mean you think nvda or tpu’s are bad, it means you see the value in lightning fast inference in CERTAIN workloads, and can sniff out bullshit in real time. Thanks Culper. Nice to see $VICR hold up well on this cbrs selloff btw. Vicor is the providing world class vertical power modules to CBRS for those unaware- this tech is one of the key reasons Cerebras is even a viable commercial product.

Yesterday, Oracle and Cerebras announced that Oracle Cloud Infrastructure (OCI) will offer Cerebras’ fast inference to AI labs, enterprises, and developers. After more than a year of extensive evaluation of Cerebras technology, Oracle selected Cerebras to bring the world's fastest AI inference to its cloud infrastructure. We're excited to bring blazingly fast inference to more AI labs, enterprises, and developers around the world. Together with Oracle.

I cross-checked the two papers Culper cites on $CBRS power draw. The numbers line up with the figures, but the papers describe measurement boundaries and operating conditions more carefully than the surrounding narrative. "Jacquemin et al. 2022 (CS-2)": The quoted 16.1 kW baseline and 22.9 kW peak are correct. Baseline is whole-chassis power (pumps, fans, power-supply losses, I/O), not the wafer alone. Coolant temperature only rose from 23.6 °C to 25.6 °C during the run, consistent with tightly managed thermals. "Golden et al. 2026 (CS-3, Llama-3.1-8B)": Idle power is shown near 80 % of TDP, and both prefill and decode run close to 100 % of TDP. Other systems in the same figure (H100, A100, MI300, Gaudi, SambaNova) sit lower during decode (roughly 45–80 %) and show lower idle. Measurement boundaries differ between board-level tools and wall power, so direct percentage comparisons need that context. From a chip designer’s perspective: large temperature swings create additional mechanical stress on a wafer-scale die and its package connections. Maintaining a more constant background load to limit thermal cycling is a reasonable option under those physical constraints. Operating at low junction temperature with cold coolant is also a different regime from the high-temperature qualification conditions typically used for wear-out lifetime estimates. The practical question is therefore economic: how the extra energy cost of a near-constant load compares with any reliability or lifetime benefit it provides. That balance depends on actual utilization. At high utilization the idle contribution becomes a smaller fraction of total energy; at lower utilization it has a larger effect on total cost of ownership. Fleet-level failure and replacement data would help quantify the trade-off. Not investment advice. Just trying to separate the engineering details from the narrative. $CBRS

Sept. 17 — GlobalFoundries and Marvell expanded their chip-capacity agreement for AI data-center optical connectivity. Reuters⁠ The multi-year deal adds SiGe capacity in Vermont for pluggable, near-packaged and co-packaged optics; volumes and financial terms were undisclosed. POV : If demand materializes, the deal could improve $MRVL supply assurance and $GFS fab utilization—further evidence that optical interconnects are becoming a core AI bottleneck.

Sept. 15 Axelera AI launched its second-generation “Europa” inference chip and secured multiple European AI-factory supply contracts. The startup reports 600+ customers and deals worth tens of millions; its $1.5B sales pipeline is prospective—not booked revenue. Comment : Europe’s sovereign-AI buildout is creating alternative inference-chip demand, with deployments flowing through $DELL and $SMCI systems.

FCC’s recent policy shifts consistently point toward LEO becoming a core communications infrastructure layer integrated with terrestrial networks, not just satellite internet. From SCS/D2D to relaxed device rules, improved spectrum efficiency, and unlicensed-device connectivity, the core LEO thesis is increasingly embedded in the regulatory roadmap. Ultimately, the real opportunity may lie not in the number of satellites, but in the NTN modem, RF, and connectivity ecosystem linking billions of devices across terrestrial and satellite networks.