Saturday, October 10, 2026

Back to today

From X

@convequity 3 clicks

Convequity on X

$NET's AI hosting is limited by the physics of its edge footprint and by the role of its products. Carrier-hotel sites built for fast network traffic are a different proposition from purpose-built AI datacenters. Cloudflare’s own engineering disclosures acknowledge power constraints, and management distinguishes its strategy from the hyperscalers’ GPU spending race. Our assessment starts with a two-H100-per-server edge configuration. H100 PCIe and NVL-class cards operate around 350–400 watts per GPU, versus up to 700 watts for H100 SXM. Hyperscalers commonly pack eight SXM GPUs into a machine and connect those machines into dense clusters. The distinction is power, cooling, memory bandwidth and GPU interconnects — not simply whether a model can run. The existing edge footprint cannot acquire hyperscaler-class cooling and power density through server upgrades. Dense, liquid-cooled Blackwell and Vera Rubin systems require a different infrastructure commitment. We see little chance that NET’s existing edge footprint becomes a competitive home for those dense clusters. For sustained, large-model inference, our assessment is that sparse edge GPUs cost more per token than dense AI clusters. Geographic proximity saves network transit time, but it does not guarantee a faster first token or a cheaper response. Greater compute capacity, faster interconnects and efficient batching can outweigh the edge’s network advantage. We expect heavy training, frontier-scale hosting and the most demanding next-generation inference to remain concentrated with hyperscalers. Workers AI’s progress with larger open models is real. Hosting those models does not establish an economic advantage over dense clusters. AI Gateway also lets customers route requests to external providers, matching the model and infrastructure to the task. Cloudflare can still win substantially. More agents mean more requests to secure, more Workers executions, more orchestration and more demand for fine-grained access controls. Those workloads build on infrastructure and capabilities Cloudflare already has. On the SaaS Survival Scorecard, we score $NET at 2.5 for Model-Layer Leverage. That reflects its substantial work building an agent harness across inference, orchestration and execution, while stopping short of model post-training and reinforcement learning.
@StableBread

Fajasy on X

Elmet $ELMT expects Masan High-Tech Materials, its Vietnamese tungsten supplier, to stay its key refiner even after a U.S.-funded plant in Nevada comes online, and it bought 4.99% of the company for $124.75M on October 1. Masan High-Tech Materials (UPCoM: MSR) mines tungsten at Nui Phao in Vietnam, refines the concentrate into APT (ammonium paratungstate, the traded form of tungsten) and oxides, and calls itself the largest producer of processed tungsten outside China. Elmet, which turns tungsten into parts for U.S. missiles and jets, has bought from MSR for 12+ years, and the deal values MSR at $2.5B. The 55.1M shares came from a Masan Group (HOSE: MSN) subsidiary, so the $124.75M goes to MSR's parent, not into MSR. Along with the stake come (1) an MSR board seat, (2) an 18-month lockup on Elmet's MSR shares, and (3) 8+ years of supply that MSR puts at ~1,250 tonnes of WO₃ (tungsten trioxide, the standard measure of contained tungsten) a year, ~$1.5B of revenue to MSR at today's prices. That's ~$187.5M a year of spend with MSR ($1.5B / 8), equal to 82% of Elmet's $228.5M of TTM revenue. On September 14, Elmet's deck labeled ~$100M of the U.S. Department of War's (DoW) $450M "Strategic Investments in Elmet Refining & Trading," its new tungsten buying and trading division, to build a "diversified feedstock portfolio." The October 1 deck relabels the same ~$100M "Strategic Investment in Masan," and the stake cost $124.75M. At $2.5B, Elmet paid 10.7x the ~$233M of net profit MSR expects for 2026 ($2.5B / $233M). For context, APT was $3,040/MTU (metric ton unit, 10 kg of WO₃) on May 29, up from $330 in January 2025 per Elmet's deck, and Elmet's own 2031 targets assume $1,500/MTU. Elmet's 8-K also says it won't participate in MSR's 2026 interim dividends. On October 1, the day the shares traded, MSR's board proposed a second 2026 interim dividend of VND 5,000/share (Vietnamese dong). VND 5,000 is 8.5% of the VND 58,760/share Elmet paid, or ~$10.6M on its 55.1M shares. In September I wrote that Springer, the Nevada APT plant, gives Elmet the one step it didn't own: Turning tungsten concentrate into the oxide its powder line starts from. Elmet's deck calls MSR a "long-time supplier and processor of blue tungsten oxide (BTO)," so MSR already handles the conversion for Elmet. So the step Elmet didn't own in July stays with MSR, and Elmet paid 10.7x MSR's expected 2026 profit for its 4.99% while its own 2031 plan assumes APT at under half the May price.

TSOH Weekly Roundup (10/09/2026)

Paramount Skydance’s debt-heavy Warner Bros. Discovery acquisition, streaming bundle pricing and alcohol-industry diversification anchor…

Boyz in the Hood

Robinhood’s AI agents, 24/7 equity trading and 10:1 crypto perpetual futures encourage higher retail trading velocity, leverage and…

Namib Minerals: Research Note

Namib Minerals posted H1 adjusted EBITDA of $19.0 million, up 76%, as higher gold prices offset lower production.

BNTC · long

Benitec Biopharma INC

Benitec Biopharma's BB-301 gene therapy is in a Phase 1b/2a trial for an ultra-rare muscular dystrophy.

Golar LNG q2’26 Operational Update

Golar LNG ordered a $2.45 billion fourth FLNG vessel, targeting 3.5MTPA of capacity and $400 million in annual EBITDA under a 20-year…

Top Rebound Stocks for Oct’26

Intuitive Surgical, down 32% from its January 2025 peak, faces slower US procedure growth and new robotic-surgery rivals despite $10.1…

Publications we follow

Every publication on this site, hand-picked.

109 publications

Newsletters 92

X accounts we follow

Every account on this site, hand-picked.

74 accounts

Investor letters we follow

189 firms

Sectors

Ordered by how often our editor reaches for them.