Permanent copy of issue 42. Today's issue is always the front page.
Nvidia said it is working with eight Australian data-center and cloud partners on land, power and shell capacity for DSX AI factories, targeting up to 2 gigawatts by 2027. Anthropic said it disrupted malicious uses of Claude between December 2025 and August 2026, including a Russia-linked espionage campaign and distillation it attributed to Alibaba, Moonshot and DeepSeek. NASA and IBM released an open-source lunar foundation model on Hugging Face, with code on GitHub.
Date: 2026-09-11
Updated: 2026-09-11 06:05
Status: LIVE
Broadcast
Raj Mirpuri, Nvidia vice president of global AI clouds, said the DSX platform is meant to host multiple generations of Nvidia compute and to expand access to Nemotron open models. Firmus said Project Southgate is aimed at hyperscale and AI-native demand across Australia and Asia-Pacific. Anthropic said Claude Haiku, Sonnet and Opus were used in the misuse cases, and that Fable and Mythos-class models were not, except in one distillation case. Kevin Murphy, NASA's acting chief data and AI officer, said the lunar model is a way to make petabytes of Moon data easier for scientists to explore.
Watch
Stories
Nvidia said it is collaborating with eight Australian data-center and cloud firms, including Firmus, Sharon AI, IREN, CDC, NEXTDC and AirTrunk, on land, power and shell capacity for DSX AI factories, targeting up to 2 gigawatts by 2027. Sharon AI said it is deploying up to 68,000 Nvidia GPUs. IREN named an 800-megawatt campus at Bundey in South Australia.
-- Two gigawatts of factory capacity is cluster power. It is not a desk.
2026-09-09 / Nvidia / energy, chip / central / #nvidia-plans-2gw-of-australian-ai-capacity
Anthropic said its Threat Intelligence team disrupted malicious uses of Claude between December 2025 and August 2026, including a Russia-linked espionage campaign whose tradecraft it said was consistent with Midnight Blizzard. The company said operators it linked to Alibaba ran more than 151 million exchanges from May to July 2026, peaking at nearly 3 million a day from more than 3,500 accounts, and that Moonshot and DeepSeek routed customer chats through Claude for training data.
-- A closed API is still a single meter someone else can copy from.
2026-09-10 / Anthropic / chip / central / #anthropic-says-labs-distilled-claude-at-scale
NASA and IBM released the NASA-IBM Lunar Foundation Model, an open-source foundation model for lunar science trained mainly on Lunar Reconnaissance Orbiter data and hosted on Hugging Face, with the codebase on GitHub. IBM said the model beat widely used methods by up to 23 percent on identifying ice deposits, craters and volcanic features, and cut ice-prospecting error by up to 22 percent versus a SwinV2-B baseline.
-- Open lunar weights are a desk-side map. They are not a training cluster.
2026-09-10 / NASA / efficiency / local / #nasa-and-ibm-release-open-lunar-model
Axis
local [####################----------] central
Score: 67
clusters still winning
up 1
Nvidia attached a 2-gigawatt Australian factory target to eight landlords. Anthropic said closed-model APIs were being stripped for training data. NASA posted lunar weights. The cluster still holds the training and the power.
Forces
| Force | Meter | Score | Pull | Note | ||
| CHIP / CAPEX |
|
84 | central | Nvidia's Australian DSX stack is cluster silicon. NASA's lunar weights are not a training cluster. | ||
| ENERGY |
|
89 | central | Nvidia named a 2-gigawatt Australian buildout by 2027. IREN's Bundey campus is 800 megawatts. | ||
| EFFICIENCY |
|
46 | local | NASA and IBM posted an open lunar model. Anthropic said other labs were copying Claude instead of training from scratch. |
Line charts (0=local 100=central 2026-08-26 to 2026-09-11)
Market overlay (2026-08-13 to 2026-09-11)
Latest: BTC $77,083 / hashrate ~934.18 EH/s / token markers avg $5.3033/M / axis 67. These lines are different units. We compare timing, not dollars-to-hashrate.
Not shown: GPU cloud $/hour, hub electricity, miner hashprice, or lab cost per token -- none of those are a clean public feed.
Power and list prices
BTC $77,083 / hashrate ~934 EH/s. List-price gap (not a subsidy %): 44.4x list gap (blended) GPT-5.6 Luna -> Claude Fable 5 ($0.45/M to $20.00/M). (mempool.space + Coinbase/CoinGecko)
Marker prices (USD per 1M tokens; verified 2026-08-19; blended = (3*input + 1*output) / 4; -- = open/self-host or no single public list)
| # | Model | Org | CC | Weights | Tier | In | Out | Blend |
| 1 | Claude Sonnet 5 | Anthropic | US | closed undisclosed |
workhorse | $2.00 | $10.00 | $4.00 |
| 2 | GPT-5.5 | OpenAI | US | closed undisclosed |
flagship | $5.00 | $30.00 | $11.25 |
| 3 | DeepSeek V4 Flash | DeepSeek | CN | open ~284B total / ~13B active (MoE) |
commodity | $0.44 | $1.32 | $0.66 |
If useful, support the broadcast.
bc1qfs3yw5qlq8sxs50crzh3g84ug27gvww9npu85n
Subscribe
$ echo you@email.com | underclassic --subscribe
Previous issue | This week | Next issue | archive
One email per issue. No tracking pixels, no filler.
home | rss | page 2 | page 3 | archive | models | subscribe | data | board | mcp | llms.txt