Father, husband,Faith, family, baseball.
Rochester, MI
Joined September 2010
- Tweets3.1K
- Following190
- Followers167
- Likes2.7K
Rich Hastings retweeted
The GB300 is the best AI computer
Two frontier labs. One accelerated computing platform. Congrats to @SpaceX and @AnthropicAI on the new compute partnership, powered by 220,000+ NVIDIA GPUs inside Colossus 1. The future of AI runs on NVIDIA.
Rich Hastings retweeted
Nvidia just figured out how to put an AI data center on the side of your house. And pay you to host it.
Each XFRA node packs 16 Blackwell RTX Pro 6000 GPUs, 4 AMD EPYC CPUs, and 3TB of RAM in a Dell PowerEdge rack mounted next to the AC condenser. The homeowner pays nothing for the hardware. They get discounted electricity and internet in exchange for letting Span tap unused capacity on their electrical panel.
This sounds insane until you look at the actual constraint blocking AI infrastructure.
Hyperscalers are not GPU-limited. Nvidia ships them on schedule. They are not capital-limited. They are sitting on hundreds of billions in capex. What they cannot get is grid interconnection. A 100MW data center requires a substation upgrade that takes 4 to 7 years in most US markets. US grid operators have over 2,600 gigawatts stuck in their interconnection queues per Lawrence Berkeley Lab. The wait, not the silicon, is the bottleneck.
Span solved this by going behind the meter.
A new Pulte home has 200A service. That's 48kW of capacity. The home uses 1 to 3kW most hours. The headroom never gets touched. Span's smart panel measures real-time consumption and dynamically routes whatever the home isn't using to the XFRA node. No substation upgrade. No queue. Just slack capacity sitting on the residential side of the meter, already cleared.
Span claims it can deploy 8,000 nodes for one fifth the cost of a comparable 100MW centralized facility, six times faster.
PulteGroup is the wedge. They delivered 29,000 homes in 2025. The XFRA unit goes in during construction next to the smart meter. No retrofit. Pulte gets a feature on the spec sheet and revenue share on the compute that flows through the wall.
The grid was the bottleneck. Pulte just became the workaround.
Rich Hastings retweeted
What US city has the largest gap between national perception and reality?
My answer: Detroit
Everyone seems to think it’s a run down hell hole w nothing but boarded up buildings.
In reality:
> Completely rejuvenated downtown w businesses, residential, restaurants, & nice sports venues – Ford Field (Lions), Comerica Park (Tigers), Little Caesars Arena (Pistons & Red Wings)
> One of the best airports in the country – Delta hub w direct flights all over the world
> Tons of history, culture, and classic architecture
> Diversifying economy beyond auto into tech, medical, etc.
> People are genuinely nice but also super tough & resilient
Of course it’s not all unicorns and rainbows. While some parts of the city are making a comeback, others are being left behind.
There’s crime, traffic and all the other things most big cities battle.
But if Detroit has a national reputation of 1/10, in actuality I think it’s more like 7/10.
Rich Hastings retweeted
CUDA 13.1 is a big upgrade
CUDA Tile basically rewires how devs talk to NVIDIA GPUs
It lowers the skill barrier, boosts efficiency, and pulls even more AI workloads into the CUDA gravity well
This is Jensen extending the moat again
TPUs and pretenders just fell further behind
Rich Hastings retweeted
AI startup Cohere found that $AMZN Trainium 1 and Trainium 2 chips were underperforming $NVDA H100 GPUs
According to an internal confidential Amazon document from July obtained by Business Insider.
Cohere reported that access to Trainium 2 was extremely limited and affected by frequent service disruptions, the document noted.
Stability AI, a startup known for generating AI images, had similar concerns. It concluded that Trainium 2 underperformed the H100 on latency, making it less competitive in both speed and cost.
An $AMZN spokesperson said the company is grateful for customer feedback that helps make its chips even better and more widely used. The Cohere case is not current, the spokesperson added, noting that Trainium and Inferentia have achieved great results with customers including Ricoh, $DDOG, and Metagenomi.
“We expect to accommodate more customers starting with Trainium 3, previewing later this year,” the spokesperson said. “We’ll accomplish that as we always do by listening to our customers and remaining vocally self-critical as we continue innovating to give more customers access to chips with the best possible price and performance.”
Last week, $AMZN CEO Andy Jassy said during the company’s earnings call that Trainium 2 chips are fully subscribed and represent a multibillion-dollar business growing 150% quarter-over-quarter. Spokespeople for Cohere and Stability AI didn’t respond to requests for comment.
Other AWS customers have also raised complaints about $AMZN AI chips. According to the July document, a startup called Typhoon found $NVDA A100 GPUs to be up to three times more cost-efficient than AWS Inferentia 2 chips for certain workloads.
Similarly, a research group called AI Singapore determined that AWS G6 servers, equipped with $NVDA GPUs, offered better cost performance than Inferentia 2 across multiple use cases. Inferentia chips are used for inference, while Trainium chips focus on training models.
A new $38 billion partnership between AWS and OpenAI highlights $AMZN ongoing challenges. The deal involves AI cloud servers that only contain $NVDA GPUs, with no mention of Trainium processors.
Rich Hastings retweeted
Nvidia announcements today:
- Eli Lilly partnership
- Palantir partnership
- Hyundai partnership
- Samsung partnership
- $1 billion investment in Nokia
- Uber partnership to build 100,000 robotaxi fleet
- $500 billion in expected revenue over through 2026
- New system connecting quantum computers to its AI chips
- Department of Energy partnership to build 7 new supercomputers
Did we miss anything?
Rich Hastings retweeted
$NVDA
This is what you are buying when you buy a share of Nvidia...
I heard this today and could not stop thinking about it.
Jensen describes the sophistication that is required in designing chips, creating them, delivering them to customers, installing them in datacenters, and more.
He argues that there has not been a company since $IBM that was faced with this much technology and investment in order to push the entire world forward with demand that is spanning throughout the globe in EVERY single industry.
The stock trades at 27 times forward earnings.
Costco and Walmart, amazing companies, are both more expensive than the name that is fueling the AI revolution and brings home 55 cents on every dollar of revenue as net income.
This has been my main Mag 7 pick for the year in terms of establishing exposure to big tech and it's just getting harder to imagine a world in which AI matters but $NVDA is not more valuable than it is today.
Rich Hastings retweeted
Jensen speaks at The White House about the $500B AI investment in the US.
It’s a good sign for $NVDA demand that Trump appears to support accelerating energy development for AI in the U.S.
Rich Hastings retweeted
NVIDIA CEO Jensen Huang stopped by the @DennysDiner mobile #GTC25 diner for breakfast🍳 trying out the special 'NVIDIA Breakast Bytes' menu item.
Huge thanks to Denny's for celebrating the GTC keynote with us and honoring Jensen's history with the company 💚
Rich Hastings retweeted
HPE annouces first shipment of NVIDIA GB200 NVL72
HPE is recognized as a leader in direct liquid cooling technology, having built seven of the top 10 world’s fastest supercomputers.
HPE delivered eight of the top 15 supercomputers on the Green500 list, which ranks the world’s most energy-efficient supercomputers.
Features of NVIDIA GB200 NVL72 by HPE:
•72 NVIDIA Blackwell GPUs and 36 NVIDIA Grace CPUsinterconnected via high-speed NVIDIA NVLink
•Up to 13.5 TB total HBM3e memory with 576 TB/sec bandwidth
•HPE direct liquid cooling technology
Rich Hastings retweeted
Number of H100 chips bought in 2024:
- $MSFT: 450,000
- $META: 350,000
- $AMZN: 196,000
- $GOOG: 169,000
If you believe they couldn’t find the way to make better AI without more chips but a few Chinese engineers did it as a side project, you are too naive.
Deepseek was reportedly trained on over 200,000 H100s.
Even if Deepseek achieved to match OpenAI with less chips, this isn’t any bearish for chip makers, to the contrary, it is amazingly bullish.
If Deepseek really reached this level with just a few thousand chips, can you imagine what could be done with a million chips?
Deepseek news, true or false, are amazingly bullish for chip makers, especially for $NVDA.
Rich Hastings retweeted
$NVDA - MUSK SUGGESTS DEEPSEEK 'OBVIOUSLY' HAS MORE NVIDIA GPUS THAN CLAIMED
Elon Musk and Alexandr Wang suggest DeepSeek has about 50,000 NVIDIA Hopper GPUs, not the 10,000 A100s they claim, due to U.S. export controls. Musk, with experience from xAI, agrees with Wang's assessment. Analysts at Cantor see this as positive for NVIDIA, despite a pre-open drop of 11% in NVIDIA's stock due to AI spending concerns.
Rich Hastings retweeted
EXCITING NEWS!!
I have been conceptually APPROVED by Michigan State University to receive my crafters license; meaning, I’m able to create these beautiful, handmade blankets, using their logo and sell them!!!
With that being said:
New pricing ✅
LLC ✅
New Blanket Patterns ✅
Rich Hastings retweeted
These kicks are WAYMO flyer than yours 😏🥶❄️ // @_trackislife94
•
•
@bigsgjamo // @Lions // @HonoluluBlues_ 🦁💙🤍🤍
Rich Hastings retweeted
Hyperscaler: Hey Nvidia, we don’t want to be locked into your ecosystem. I’ll go to AVGO or MRVL to develop custom chips.
AI Developers: That’s fine for you, but I don’t want to be locked into your ecosystem either. I prefer Nvidia because then my software works everywhere, from cloud to edge.
Hyperscaler: But my chips will be cheaper to rent.
AI Developers: Is that really true? Your future chips might be cheaper than Nvidia’s current offerings, but they could end up more expensive than Nvidia’s future offerings.
Hyperscaler: Okay, no problem. I’ll still buy Nvidia chips so you can rent them through my cloud. Just don’t go elsewhere.
AI Developers: What are you using your chips for, then?
Hyperscaler: Mostly for my internal workloads. Still, I’ll keep buying Nvidia chips for as far into the future as I can see.
Jensen: Hold my beer.