Anthropic Signs $35 Billion Cloud Deal With Nvidia-Backed Lambda (reuters.com) 18
Anthropic has reportedly signed a $35 billion cloud-computing deal with Nvidia-backed Lambda for a roughly 350-megawatt Texas data center being developed by Hut 8. "Lambda will install Nvidia chips at the Hut 8 facility, with the arrangement designed to expand the pool of Nvidia computing resources available to power Anthropic's Claude AI products," reports Quartz. From the report: The arrangement is a further sign of how Nvidia is positioning itself as a facilitator of computing access across the AI industry. It also benefited Lambda, which counts Nvidia as an investor and was able to land a major Anthropic contract without having to independently source or lease the underlying data center space, according to the Journal.
The data center covers roughly 350 megawatts of capacity, according to Bloomberg. Hut 8, which began as a bitcoin mining company before moving into AI data centers, disclosed in July that an investment-grade customer had signed 15-year leases for its Texas campus, with a base contract value of $19.6 billion, without naming the tenant.
The Lambda deal is the latest in a series of large computing agreements Anthropic has pursued. Last week, the company agreed to spend $45 billion to rent capacity from Nscale's West Virginia data center campus. It has also signed deals worth $50 billion with neocloud Fluidstack Ltd. and $45 billion with SpaceX, according to Bloomberg.
The data center covers roughly 350 megawatts of capacity, according to Bloomberg. Hut 8, which began as a bitcoin mining company before moving into AI data centers, disclosed in July that an investment-grade customer had signed 15-year leases for its Texas campus, with a base contract value of $19.6 billion, without naming the tenant.
The Lambda deal is the latest in a series of large computing agreements Anthropic has pursued. Last week, the company agreed to spend $45 billion to rent capacity from Nscale's West Virginia data center campus. It has also signed deals worth $50 billion with neocloud Fluidstack Ltd. and $45 billion with SpaceX, according to Bloomberg.
Another Former Mining Company (Score:4, Insightful)
Lots of low cost electrical power (Score:2)
With so much computing power and focus pivoting from cryptocurrency to LLM operations, I can't help but wonder if Bitcoin and the gang's best days are already behind them.
Its not about computing power, its about electrical power. Crypto mining is largely done by highly customized task specific ASIC hardware. This hardware can do nothing other than crypto mining. Once a bitcoin miner is no longer profitable it has not alternative use other than improvised boat anchor, its salvage value is the metal scap value of its heat sink.
At the ankle biters end of crypto mining there is some GPU mining, but all the somewhat serious stuff has moved on to ASIC long ago.
The real const
Demand Chain Crash (Score:5, Informative)
The whole data center AI investment thing looks like it is going to BUST. The latest open source models that came out this month have show 10x-100x improvements in efficiency over the last 9 months. And innovations like Deepseek Sparse Attention and other training innovations have made it clear that our CURRENT phones and CURRENT graphics cards are MORE than enough to run MAXIMALLY DANGEROUS allowable AIs already. By the end of THIS QUARTER, algorithmic efficiency alone will SHATTER the current AI investment story.
Think about it... the top models are already restricted for safety. We are now getting Qwen 27b models that are punching at 90% of frontier, and even the AMAZING little Ornith 9b model that came out a couple weeks ago that is equivalent to ChatGPT4 from a couple years back.
Why the hell do we need so much CLAUDE when 90% of people will NEVER use the AI more than casually? Even if AI usage grows.... there is NOT enough demand that will pop up this quarter to close the gap.
We are heading for a BLACK SWAN crash in days because the AI investment story is VAPORIZING before our eyes due to algo efficiency, which is wiping out the hardware investment story upon which the current cycle is based. We have a $2T house of cards going on here. And NOBODY knows how to use the 10x-100X ADDITIONAL efficiency we are likely to see from the already-discovered four or five major algo innovations that have be cross-pollinated the last few weeks, each of which has about 50-100% additional runway of efficiency. So the compounding efficiency means our already-owned PHONES will be enough to satisfy our AI needs THIS QUARTER.
Re: (Score:1)
There's quite a bit of hyperbole and assumption in what you're saying - even from enthusiasts getting Qwen model running locally "90%" of frontier is a bold statement given the capabilities of actual frontier models. I don't think you're pricing in demand - would demand decrease from cheaper models or increase? You seem to imply that we're at a saturation point from demand, but I feel this couldn't be further from the truth. There's a lot of integration points and a lot of huge benefits a 100x model coul
Re: (Score:2)
I think his point was our current hardware is becoming good enough to run more then good enough local models and we haven't even updated our hardware yet. I can't wait to see what I can run locally next year since I still don't want to pluck down money for a 16gb nvidia card, let alone a 32gb card. It's just a hobby for me and barely that. Certainly not worth spending 4 figures over.
And which AI model are you? (Score:2)
Because unless the author is mentally unwell that incoherent rant was written by machine.
Webvan signs 40-year lease on mega HQ (Score:2)
Are they going to pay for power plant construction (Score:3)
Since this is another data center that will be sucking up power from the grid, and force spending on power infrastructure upgrades, then will Anthropic be paying for a new power plant and for all the infrastructure upgrades needed for this data center? If not, they should be forced to do that before the data center is allowed to power up.
70% of RAM from The Big 3 is sold out until 2031 (Score:2)
That's about 8 years. You can no longer make the argument that nobody is entering the ram market because they're worried the bubble will pop. Everyone knew from the get-go it wasn't going to pop anytime soon and that this would be a permanent increase in demand. Or at least long enough to build a successful business off of.
Given that you have to start asking why nobody bothered entering the
I have never said "Oh, Claude Code is so fast!" (Score:2)
Computing power is not measured in watts (Score:1)
Gentle reminder to idiots at the back. Computing power is not measured in watts. DC watts are your costs of running the kit. 1MW hydro plant will generate zero tokens for Scam A.
Hands (Score:1)