← Back to leaderboard
benitoz

Ben Pouladian ✈️ GTC Berlin 🇩🇪

Verifiedactive
@benitoz

AI Infrastructure • Compute Economics • Semiconductors • Models • Power • $NVDA since 2016

RosterStartup & SaaS BuildersTrading & MarketsAI Model News & Releases
25,702 followers · 1,311 following · Last polled 3h ago
Perf. Score
44.3
Rank #274
Value Score
22.5
#217 of 273 priced
Median views
3.5K
p25 2.5K · trailing 7d
Followers
26K
+0.7% 7d
ER (impr.)
0.81%
eng ÷ impressions
ER (foll.)
0.41%
eng ÷ followers
Est. CPM
$98.84
QT-rate basis
WoW views
+188.5%
▲ rising

Rates & value · est. from 3.5K median organic views

overpriced
Quote tweetbasis
$350
$98.84 / 1K views
Post
—
no CPM
Thread
—
no CPM
Retweet
—
no CPM
Value Score 22.5 · rank #217 of 273 priced$8139.53 / 1K engagementsrate age unknown (pre-audit import)

Trends

Follower growth
Median views / day
Engagement rate
View distribution · trailing-7d posts

Each blue dot is one organic post. The Performance Score ranks by the median (3.5K), not the mean (13K) — so a single viral post doesn't inflate a creator's reach. The p25 floor (2.5K) is the level 75% of posts clear.

Recent posts · ranked by views

PostPostedViewsLikesRepostsRepliesQuotesBookmarksER
. @JensenHuang came to America as a child and worked his way to the top. Now, the National Medal of Science, with his parents there to share the moment. As a first-generation American whose parents emigrated from Iran for a better life, this hits home. They believed in America as a land of opportunity. Imagine how proud his parents must be. 🇺🇸 $NVDAfrozen14d ago117K5.0K479147432074.99%
Jensen handled it with grace and class He’s truly a generational leader and entrepreneur 💚 https://t.co/dvxvIh9UpE1d ago114K373151341630.50%
For my local Ai brothers what about connecting all your DGX Sparks? 200 GBs switch Coming soon 👀 https://t.co/WvFlivxhzn
1d ago
71K
282
28
44
32
85
0.66%
When the wealthiest man in the universe is posting memes about trying to rent AI compute… Pay attention, anon 🤣 $NVDA $SPCXfrozen4d ago58K3411891750.76%
Everyone's watching GPUs. Jensen's watching substrate. "Substrates are getting larger and there are more chips... it's going to be hard to close that gap for a few years." Morgan Stanley's Oct 14 ABF note: undersupply hits 9.3% in CY27 and 13.7% in CY28, prices rise 25%+, and new capacity doesn't arrive until 2028 to 2030. Aunty Lisa took her trips to Taiwan too late. The real bottleneck is the thing the chips sit on anon $NVDA $AMDfrozen4d ago46K1301975510.47%
"Sight unseen, who can you place a $50, $100 billion order on? We're literally the only one." - Jensen Hyperscalers order racks a year or two before they see them. That's the bar Helios has to clear. I was at Moscone for AMD Advancing AI in July. Helios looked credible on paper, but it's AMD's first rack-scale system and still unproven at volume manufacturing. My call was meaningful volume in 2H27, two quarters past AMD's own window. Meanwhile AMD is getting squeezed from the other side by the custom inference silicon showing up everywhere. Nobody writes a $50B check on a slide. https://t.co/Qv6GSg3QIc $NVDA $AMDfrozen4d ago31K11014112530.62%
Here is the Microduck up close You can program it work DGX spark This baby duck represents the future of physical Ai it relies on reinforcement learning to transition AI-trained movements from digital simulations into the real world Small duck now large robots later 🐥🤖💚 https://t.co/a6HlAwJJez1d ago26K29135234581.59%
Elon’s point is the memory wall: once the workload fits in memory, feeding the chip can become the bottleneck. Tesla says it cut RAM capacity while preserving bandwidth. Our January breakdown of why bandwidth matters: https://t.co/IopwOoZuN1frozen21d ago21K57521320.46%
Talent and access to compute. Poolside’s licensing deal. The agreement to buy Hugging Face. Now Reflection talks. If this closes, it should strengthen Nemotron and America’s ability to rival China in open models. We wrote it September 3: “The Only Company That Can Afford Free.” NVIDIA can fund the talent and give away the models because adoption sells compute. Influence through inference. https://t.co/jw3OpiNIcVfrozen12d ago18K15617112341.21%
Gavin engaging in a healthy debate with Ai bear/skeptic Ed Zitron who for some reason has blocked me. I’ve never interacted with the dudefrozen9d ago17K37090140.36%
Nobody's sitting through a 1.5-hour podcast, so a 10-second clip becomes the whole story. That's not fair to the person being interviewed. Here's the full 3 minutes of the viral Jensen "long division doesn't matter" moment. His point isn't "kids should be dumb." It's that the skill moves up a level. He knew every one of the 200 transistors on his first chip by name. No engineer does that today, and chips are better for it. You don't need to know how every resistor in an iPhone works to build a great app. You need to see the system. I'm teaching my kids the same way. They're still learning their math, but the goal is for them to see the forest through the trees: the bigger picture, systems-level thinking.frozen25d ago16K179241911122.04%
AI keeps running into atoms. NVIDIA targets SPIL CPO optical-engine 200GBps made by TSMC Coupe NVIDIA's side of the Optical Cold War: secure the optical supply behind the GPUs. The moat reaches into the supply chain. $NVDA https://t.co/Zv7ihrKYRx1d ago16K110635270.94%
Confirmed with my folks at Nvidia 6.1 Sol Ultra runs on the big green not Cerebras $NVDA $CBRS https://t.co/WJ8wZG8cCYfrozen23d ago16K13774080.97%
Trump saw Dario on SNL last night and he’s like “I got to meet this guy, set him straight” 😀 https://t.co/m81iLXcjrmfrozen25d ago13K4003030.36%
SemiAnalysis is treating GTC like a birthday party where everyone deserves an invite 🥳😭. NVIDIA is building a business. Prioritizing partners who build on its platform is hardly shocking. The big picture is customer success and platform economics. Wouldn’t you back your own ecosystem?frozen6d ago13K1157160101.17%
Buying GPUs gets the headlines. Making them productive is the work. Meta reports 99% fewer training interruptions at Prometheus from the initial peak, with NVIDIA engineers helping tackle firmware, memory and reliability issues. Meta also reports over 85% pre-training goodput and over 90% for reinforcement learning on Muse Wonder. Goodput means training time that actually moves the model forward. Our takeaway: the value is in the whole system. Chips, networking, software and recovery all determine how much useful work your compute delivers. This is what making compute count looks like.frozen3d ago12K1001163321.25%
“Same rack, different multiples.” Revisiting @GavinSBaker’s point. Since July 28: MU +29.7% NVDA +21.3% VRT −5.9% The cheaper parts of the AI rack have outperformed. So far, that relative-value call is aging well.frozen16d ago11K55220150.69%
Cheers to the weekend! These coasters came in handy for the office @tracyalloway @TheStalwart https://t.co/vMctOrEGZLfrozen27d ago10.0K1312000.16%
The “NVIDIA killer” business is turning into the NVIDIA partner program. The Information reports NVIDIA plans to invest in d-Matrix, which is already joining NVLink Fusion. Compete on the accelerator, plug into Jensen’s infrastructure. That’s the beauty of owning the platform. Even rival chips can expand your footprint. All roads lead to NVIDIA. $NVDAfrozen14d ago9.6K961053151.34%
Jensen to the CPO boys It’s coming. It took 5 years to invent the process technology. TSMC will deliver Volume is about to ramp Laser go pew pew Nvidia GTC Berlin $NVDA https://t.co/jdvdwb7JRp10h ago9.6K110761211.52%
Jensen on Ezra Klein: today's engineers are "better systems thinkers" than he was. On his first chip he knew all 200 transistors by name. Nobody does that anymore, and that's fine. The same lesson shows up in two places. AI supply chain: everyone's chasing the next bottleneck. HBM, CoWoS, transformers, power. Who cares? Fix one and the next one shows up. The people who win see the whole factory and how the pieces interact. My kids: AI tools will handle the parts for them. So I'm teaching them to think in systems. What connects to what, what breaks what, and what happens next. That's the skill the tools won't do for them. What matters is the system, not the parts. Think in systemsfrozen28d ago9.4K79880281.31%
AMD is my pick for the most overhyped chip stock. The AI trade needs receipts: shipped racks, competitive token costs and real power. I made the case with @JackFarley96 on Monetary Matters. https://t.co/wWaiwTXZT0 $AMD $NVDAfrozen21d ago9.2K372121120.70%
From a GPU to an entire data center in 28 seconds. We built this in Blender using NVIDIA’s public hardware references, then brought it into BEP Research’s interactive data-center explorer. Follow the hardware. Understand the business. https://t.co/0uG9SO0mAz https://t.co/cMri7YUxoMfrozen19d ago8.6K12015101662.45%
Some AI bubble. Lumentum CEO in April: “basically sold out” through 2027. Now: “almost sold out for 2029.” Under-shipping demand by 30% on some products into 2028; ~70% on others in 2027. Demand keeps outrunning the fabs. Laser go pew pew. $LITE https://t.co/Q45WAIwCg6frozen14d ago8.4K596134151.16%
OpenAI’s Astra Ultrafast runs on Blackwell, with AI helping optimize the kernels underneath it. Better models squeezing more performance from deployed hardware. That’s the $NVDA software moat compounding. Cerebras doesn’t own the word “ultrafast.” Blog link https://t.co/G8iRObIrnKfrozen21d ago8.2K3855170.68%
Having fun this Friday. Will delete later Read up on Zero Flaps https://t.co/bwH9xjJKLV $CRDO https://t.co/9kjQBYfop8frozen27d ago7.9K2149150.51%
The Financial Times having to clarify gross vs. net revenue is quite something for a publication with “Financial” in the name. $70B gross. $50B net. Both can be true. Before framing a $20B shortfall, maybe check whether you’re comparing the same number. Accounting 101. https://t.co/OLlq6h0utnfrozen14d ago7.9K18513252133.03%
Gas turbines make AC. Bloom makes DC. What changes on the way to an AI rack? We animated the conventional turbine path versus Bloom’s proposed 800V DC design. See which stages disappear and where conversion remains. Our research: https://t.co/APdRjpKgjz https://t.co/SxiuPz4Ymhfrozen18d ago7.8K18100120.40%
Wall Street bought CPUs. My agent still needed a human. A billion users isn't a hardware spec. We modeled the workload, the token bill and who gets paid to let agents through. Meta's Muse rally may have bought the wrong layer. https://t.co/pukDttoS71frozen1mo ago7.7K34230260.85%
Very refreshing to see a CEO like Bill Brennan of Credo know so much about his product and technology while working his booth. $CRDO https://t.co/cuPE0U2ZTefrozen8d ago7.4K64222141.14%
"It sounded like the ravings of a madman at the time." A couple years ago Jensen said AI infrastructure was headed to $3-4T a year by the end of the decade. Now the top five hyperscalers alone are on track for ~$1.5T next year. Great interview with @Srasgon on @JackFarley96's Monetary Matters: https://t.co/lxJZW2WMPpfrozen11d ago7.3K4051160.73%
It’s a rich mans game with scaling laws holding $100-$200b capex on good just vibes! $META https://t.co/5daLhMk7rZfrozen10d ago7.0K5012290.91%
My brother from Boston Ben Affleck is pro open source and open models. He knows his stuff Proud of you https://t.co/r2bvathE6mfrozen21d ago6.8K85562231.77%
AI infrastructure gets a lot more interesting when you follow the data. Our BEP Research team built this visual journey from the compute tray to the rack, through a massive data center and out to other sites. Lasers carry the bits. Tiny mirrors switch the optical paths. This is why we think at the system level, and why companies like Lumentum matter. $LITE Sound on 🔊 Take 103 seconds and see how it connects.frozen11d ago6.6K23602120.66%
FT’s ~$50B OpenAI run rate is a September snapshot. @FundaAI has it at $48.8B, close to that headline. Dots launched Sept 29. @thsottiaux now reports 40M active users across Codex + Work. BEP’s hypothesis: dots, new models and tiers, plus coding and enterprise adoption, could drive $100B by year-end. That requires ~26% monthly growth on the same revenue basis. Ambitious, not confirmed. As @altcap argued at All-In, downstream revenue must justify upstream capex.frozen14d ago6.5K3955190.91%
"One of these days people might not categorize us as a chip company anymore." - Jensen Most investors still model Nvidia as a GPU company. Meanwhile he's laying out a software licensing business: a couple thousand dollars per GPU per year, across millions of GPUs a year. I wonder how many investors know what CUDA-X is. Or Dynamo. Or Nemotron. CUDA-X is the library layer whole industries build on. Dynamo is the open source inference stack that runs entire GPU fleets. Nemotron is Nvidia's own family of open models. The chip is the wedge. The stack is the business. We'll dig deeper into this at BEP Research and @NVIDIAGTC Berlin. $NVDAfrozen6d ago5.7K3763080.94%
Meta just hired MongoDB's CEO, CJ Desai, to run a new Meta Enterprise Platform. Most takes read it as a run at the CIO. I think the customer that matters is a shop owner in New Delhi who already runs the business on WhatsApp. Meta says over a million businesses already use its Business Agent, and it's on Desai's launch list. Back in April we wrote: "WhatsApp is not competing with Microsoft." I finally started a $META position today. I took the paywall off: https://t.co/Lw2kaTT9rifrozen24d ago5.5K1321260.44%
“AI-yatollah” was too good not to turn into a poster. 😂 Introducing the Supreme Leader of Artificial Intelligence. For your safety, of course. https://t.co/QBEsh3s1hrfrozen24d ago5.5K4376041.10%
A wafer can work and still miss the performance spec. That’s the parametric-yield question we’ve been digging into at BEP. For Cerebras, speed has to survive manufacturing, repair and testing at scale. Useful work from @damnang2 on those trade-offs. Modeled scenarios, not disclosed production yields.frozen14d ago5.4K16231130.64%
Always a great time to hang out with @JensenHuang at @NVIDIAGTC Berlin! I’m not ambitious enough! 😄 🇺🇸🇩🇪🚀🤖💚 https://t.co/RsknCaLp7p13h ago5.4K13738142.85%
Caught this live on @jackfarley96's Monetary Matters yesterday. Jack asked which names in the AI science basket had real explosive upside. I said Twist: go on their site, program a cell, programmable biology. Anthropic partnership already rerating it. Claude in the loop. They just opened a wet lab in SF. This is not Claude code anymore. We're doing physical sciences. $TWST ripping today on the Google news. Still early.frozen22d ago5.2K1314060.46%
Remove the bottleneck An unplugged GPU produces zero tokens. NVIDIA says DSX POD cuts installation from up to 6 months to 14 days after site readiness. The modular approach makes time to first token faster and less on site labor. $NVDA: engineer the whole factory and get it producing sooner.1d ago5.0K47420171.40%
JENSEN IS LOCKED IN 🔒 FOLLOW HIS LEAD BE MORE AMBITIOUS AI IS THE GREAT EQUALIZER https://t.co/GPNdaJh591frozen29d ago4.8K2254140.75%
When @DavidSacks or @chamath says EA this is what he's talking about, don't let them slow us down @Jason !frozen27d ago4.8K711010.21%
Anthropic's $42B loss is mostly non-cash. Per Reuters' read of the prospectus, about $34B is a markup on financing that can convert to shares 2025 compute was $7.33B on ~$4.6B of revenue Q1 2026 revenue alone was $4.8B (WSJ, unaudited) The IPO prices the Compute Crossover https://t.co/dEtP41F1uyfrozen24d ago4.7K26211040.92%
I tried pricing Elon's GPU list the way a Nebius customer would, and it came down to something boring. Who buys the spare hours. 780k GPUs at Nebius's own preemptible rates is ~$25B a year if every hour sells. ~$19B at 75%. The spot floor gets you $6B. On-demand, $46B. Anthropic already pays SpaceX $1.25B a month. Another 220k GB300s land next week. Nebius raises on-demand about 20% on Oct 1. Decent side business for a rocket company, no?frozen28d ago4.4K2122060.71%
new deep dive from the Peter Lynch of Portugal Manuel, take a look and sub, interesting company and a space that I think is perfect for applied AIfrozen27d ago4.1K612010.24%
Called it Apple crushed it and people already selling Pre-Orders of the Duo on Ebay $AAPL https://t.co/C39gOgFLlUfrozen6d ago4.1K601100.20%
Excited to see what Nick @Midnight_Captl , @stevehou , and @carmenli do for the compute markets! LFG 🚀🤖💚frozen24d ago4.0K1401010.40%
Can’t wait for the Costco MCP Throw me a party. Make no mistakes https://t.co/MBtkSQg9lbfrozen19d ago4.0K2402030.73%
Claude proposed the experiments. Humans ran them Our AI x biotech basket owns the lab gate Street hears invent We own Make and Test The Moat Is the Measurement https://t.co/1PPzLE1foC The next inflection is the labfrozen29d ago4.0K833030.43%
Jensen’s ~$2,000 per GPU per year in enterprise software deserves more attention. Enterprises deploying AI agents need security, governance and someone accountable for keeping the stack running. My bet: OpenShell becomes the standard security layer for enterprise AI, driving adoption of NVIDIA’s paid software and support. Consider the math. Assume: • 10M eligible enterprise GPUs shipped in a year • 50% paid software attach • $2,000 per GPU annually That’s $10B in annual recurring revenue potential. Keep those subscriptions for five years and you get $50B from a single year’s GPU shipments. Each new year of GPU sales adds another potential stream of software revenue. $NVDA9h ago3.9K3333181.23%
True: Agents will make good recommendations False: they will actually move money for you https://t.co/CzPmcO9IgYfrozen25d ago3.8K623000.29%
This is the NY–SF disconnect in one debate. One side sees a falling AI bill. The other sees another workflow becoming economical. I’m with @GavinSBaker on the mechanism. Cheaper intelligence can expand the market. We laid out the investment question at BEP after DevDay: does the extra work outrun the efficiency gains? That’s what connects agents to infrastructure demand. We’ve been building the task-cost tools to test it. https://t.co/eYnadMI2Oefrozen13d ago3.8K25430111.13%
Made it to Berlin for @NVIDIAGTC 🇩🇪 Excited to see how Germany’s Mittelstand is embracing physical AI and what’s ahead for robotics, manufacturing, and industrial automation. It’s time to build 🤖💚 Big thanks to NVIDIA for hosting me at GTC Berlin 🍻 https://t.co/ASZi33HAWT2d ago3.8K80211002.46%
I keep coming back to Helios for the memory, not the flops. AMD's look-back counted parts: 18,000+ compute units, 4,600 CPU cores, 31 TB of HBM4 in one rack. In July I photographed AMD's comparison slide from my seat at Moscone. Helios led Vera Rubin by 15% at low interactivity, 10% at high. The faster the answer, the more the chip waits on memory. That slide put the bandwidth edge at 6%. The lead narrows right where agent work is headed. Curious where it settles once racks ship.frozen25d ago3.8K1421020.50%
Here is the Ai Science basket weights and performance since August 10 inception https://t.co/1PPzLE1foC Past performance is not indicative of future results https://t.co/GpotdW02Xufrozen18d ago3.7K801130.35%
This is criminal Miguel, come on, do better https://t.co/JeDWe1UcLTfrozen21d ago3.7K2215110.81%
Nice Day for the Ai Science basket glad to see $ILMN getting the recognition it deserves 🤖🧬🚀 https://t.co/pJnhzWvzEDfrozen17d ago3.7K1611050.63%
In Jensen math that’s $1T of Ai capex Alone But Elon and @spacexai don’t build luxury data centers so it’s a little less call it $800B 😄2d ago3.6K1614000.58%
AI agents need computers to run their code, test their work and finish the job. We made this to show what that looks like, from the sandbox to the data center. Part 1 of our BEP Research CPU series is live and free: https://t.co/nQfbaIQE9P https://t.co/sNpCfEV9yCfrozen10d ago3.6K1831070.82%
Weekends are for philosophy, even if it’s Friday night. I spent this week watching people build the future, then opened X and apparently the whole thing depends on two AI labs and one CoreWeave debt tranche. At DigitalOcean, I saw the energy around agents. At SEMICON West, I saw the semiconductor manufacturing ecosystem working on how to make things faster and better. At OCP, it was a madhouse. Packed booths. Companies showing the power, cooling, memory, networking and systems needed to make all this work. Then I come online and the entire debate is one ARR number or a bond trading at a 15% yield. Financing matters. But bring it back to first principles: what useful work can this technology do, what does it cost, and how many people have even started using it? We already had an enormous computing industry before ChatGPT. Enterprises were already buying servers, running software and replacing aging infrastructure. Now the question is how much of that work can be done differently—and what becomes possible that wasn’t economical before. This goes well beyond a handful of frontier labs. It reaches into engineering, manufacturing, scientific research and the everyday work inside businesses. And adoption? I just got my wife using her first AI tool, Muse. Spend enough time in this bubble and you forget how many people are still at the beginning. If the United States makes infrastructure too difficult to build, other countries will compete for that investment. Japan. Australia. Europe. Power, permits and connectivity still matter, but the ambition isn’t confined to Silicon Valley. Next stop for me is GTC Berlin, at NVIDIA’s invitation, to see that next chapter firsthand. I think countries that fail to build the capacity to use AI will fall behind. That means compute, energy, skills and businesses actually putting the technology to work. None of this guarantees every company succeeds or every dollar invested earns a return. A computing transition can be enormous and still have bankruptcies, overbuilding and bad investments along the way. But judging the whole thing through a few companies’ financial headlines misses what’s happening underneath. Get out of the feed. Go see what people are building. Then ask how much of the world has even started using it.frozen6d ago3.5K2638151.21%
Everyone dissecting CoreWeave’s debt should bring the same microscope to Anthropic. CoreWeave’s widely discussed bond touched a 15% market yield. Its contractual coupon is 9.625%. FT reports Blackstone was offered a 10% coupon, with a higher all-in yield, on a $9B junior slice of Anthropic’s proposed TPU financing. Different structures. Different risks. But switching the chip logo doesn’t make the financing bill disappear. The question for both: how much cash does the compute generate after paying for the full stack?frozen5d ago3.5K1427150.82%
Steve Jobs said you can only connect the dots looking backward. Windows → GPUs → CUDA → AI. We put together a visual journey through the breakthroughs and lucky turns that connected NVIDIA, Microsoft and OpenAI. @JensenHuang @satyanadella @sriramk https://t.co/gpVkpYWbdjfrozen15d ago3.5K41320101.59%
Wow. Loving my dot I can just call it and get stuff done Nicely done OpenAI team The keyboard is now optional https://t.co/9EyFfh4B5Mfrozen20d ago3.5K2815021.04%
If you’ve followed our $NVDA work, tomorrow should connect a few dots. The factory is the product. The moat reaches well beyond the GPU. Watch how far Jensen takes that at Nvidia GTC Berlin. More tomorrow. 🤖💚 https://t.co/QyQkX8d11K2d ago3.4K5441041.87%
What a wild day 2 at @NVIDIAGTC Berlin I got the chance to sit down and learn more about @silicondata @computeexchange from @carmenli More later at BEP Research https://t.co/LGJNEWhthH7h ago3.4K2821110.98%
Jensen just dunked on Jordan. 5 years: $NVDA +1,013%, $NKE -76%. Tonight Nike guided FY27 EPS to $1.15-$1.35 vs ~$1.67 Street. That's 28x earnings for a company guiding sales down. NVIDIA trades at 25x. The shoe now costs more than the chip. Just don't. https://t.co/9IVYoB9xGWfrozen21d ago3.4K3026021.19%
As I head to Europe for GTC Berlin, I keep coming back to a question: what good is building the future if our children don’t have a stake in it? As a father of three, this is personal. I want my kids to grow up knowing they can help build something and own a piece of it. But that opportunity shouldn’t depend on whether their parents already have money. Watching the unrest in France, I worry about what happens when people lose faith that the system has a place for them. Ownership won’t solve every grievance. But I believe capitalism is stronger when more people participate in its upside, rather than watch it from the outside. That’s why @altcap work with Invest America, @MichaelDell and Susan Dell’s gifts, and Elon’s pledge here deserve real credit. Different efforts, united by a powerful idea: give children a stake in the future before they’re old enough to feel excluded from it. A first investment can teach something a lecture cannot. You own a small piece of a business. You can watch it grow, weather setbacks and learn what patience means. Compounding becomes something you experience, not just something you’re told about. Everyone wants equity. Not just a seat at the table, but a share in what gets built. Let’s start the kids young. 👶💪🚀🇺🇸frozen3d ago3.3K3523041.32%
Software will flow like water. Almost free to create. Apps, games, entire experiences generated around what you want. My bet: all new software becomes generative, and human taste becomes the scarce resource. What is useful? What is fun? What is worth making? Feels like another nail in the coffin for selling undifferentiated features. Big love to @matthewberman for pushing this conversation forward 👏frozen12d ago3.3K2242161.06%
This lineup keeps getting better 🔥 Jensen Huang, Lip-Bu Tan, Brad Gerstner and more. I’ll be at TomorrowX in Austin, Nov 17–18. Come talk AI, chips and the business behind it all. Join me: https://t.co/HOvuLapxRv Complimentary pass; I earn referral perks.frozen16d ago3.3K2624021.04%
NVIDIA says its AI factories are productive, durable and fungible. A100 GPUs, launched in 2020, are still in commercial service six years later. In this clip from Monetary Matters with @JackFarley96 I compare NVIDIA GPUs to Fifth Avenue real estate: who wants the next lease? Our test: paying demand, utilization and cost. $NVDAfrozen21d ago3.2K1624120.79%
In April I wrote Bloom's fuel cells put out 800V DC, "no conversion required." Too clean. So we mapped the path to the rack from Bloom's Sep 16 report and our Sep 17 call. The turbine path needs a transformer, AC switchgear and an AC-to-DC stage. Bloom's skips all three. A DC-DC regulator stays. Bloom models a $3.9B edge. That's against grid-fed 800V DC. The turbine case isn't priced. $BE https://t.co/mIZn9FNubzfrozen28d ago3.1K1100130.48%
Cheaper AI can still mean a bigger compute bill. Persistent agents can add planning, tool calls and retries. Our DevDay note asks what that means for NVIDIA—and what evidence the bull case still needs. I'm long $NVDA. Free to read: https://t.co/tZ24TT9XNEfrozen23d ago3.1K920020.42%
SpaceX is looking to raise $40B to buy NVIDIA chips in an Apollo-led financing, per FT. We wrote the thesis in August. Now look at the size of the check they’re trying to write. All Roads and Rockets Lead to NVIDIA. https://t.co/cXn91MsveQfrozen16d ago3.1K4863021.91%
We laid out our AI biology thesis back on August 10. Nice to see more people connecting the dots. Cheaper AI-generated designs mean more experiments. Someone still has to supply the instruments, DNA and reagents. We’re getting to know this space better and working on more exciting research. More to come. The 20-name research basket: $BRKR $TMO $DHR $ILMN $A $TXG $TECN.SW $NVDA $SDGR $CERT $TWST $RGEN $SRT.DE $TECH $AVTR $QGEN $LLY $NTRA $TEM $ABCL https://t.co/1ZIcQz0eKdfrozen17d ago3.0K11230110.90%
looks like Meta (Zuck) is taking a data privacy page from Apple marketing $meta $aapl https://t.co/FMK2lOreNbfrozen29d ago2.9K1602010.67%
We’re tracking the same tension at BEP Research. @OrnnExchange B200 rental benchmark rose ~23% from May 31 to Oct 10, while H100 fell ~9%. Cheaper intelligence can expand demand for the infrastructure behind it. But which hardware you own matters. The physical bottlenecks still matter. https://t.co/To0xfldvW8frozen11d ago2.8K1711050.85%
“America’s SiC bet.” We wrote that last October. Today: a conditional $1.5B, 30-year DoW loan commitment, with $600M aimed at refinancing the expensive debt. Different vehicle. Same strategic logic. Nice to see this taking shape. $WOLFfrozen15d ago2.8K1402010.61%
Good point from @GavinSBaker: OpenAI and Anthropic’s revenue don’t capture the whole AI economy. Our BEP angle: open models still need GPUs, memory, networking and power. Cheaper intelligence can mean more workloads running on someone else’s infrastructure. Open weights. Paid compute. That’s the opportunity for the hyperscalers.frozen7d ago2.8K18411101.23%
The factory is the product. Compute. Networking. Storage. Software. Deployment. Our $NVDA thesis: the advantage compounds across the system. A rival has to beat the cost of getting useful work done across that entire stack. https://t.co/AmOFWNgrMg2d ago2.7K3451171.78%
Jensen stops by one of the leading companies in Germany, @Siemens to show some GTC Love 💚 Every company will be an Ai company https://t.co/ezcXKhPIPm10h ago2.7K2800051.23%
Weekends are for philosophy ☕️ The smarter the agent, the more important the question: who gave it permission? Microsoft increasingly looks like an enterprise harness company. Models supply intelligence. The harness governs what that intelligence can actually do. We’ve been writing about this at BEP since January’s “Verification Gap”: permissions, audit trails, containment and who owns the control layer. Satya’s piece brings that thesis into focus. Who captures the economics? https://t.co/cXZRqEvbZwfrozen12d ago2.7K1914041.05%
Weekends are for philosophy: Silicon Valley spent two decades telling everyone religion was irrational. Then it built a god out of matrix multiplication, appointed itself the priesthood, and started collecting $500B a year in tithes called capex. The accelerationists think it's the Second Coming. The doomers think it's the Book of Revelation. Both are certain. Neither can tell you what's happening inside the weights. Faith is fine. Just don't confuse it with a thesis Enjoy the sabbathfrozen27d ago2.6K302000.19%
Akram raises a fair question on the valuation gap. I’d add one distinction from our BEP work: Lumentum’s moat depends on delivering laser power, narrow linewidth and reliability together at scale. HBM has real qualification barriers too. The question is how durable each advantage is, and how much of it is already in the price. Same AI tailwind, different competitive dynamics. $LITE $MUfrozen12d ago2.6K902160.69%
Happy birthday @kimmonismus GTC Berlin Edition https://t.co/H9P7blqXsE1d ago2.6K1111010.54%
The calm before the storm at GTC Berlin. @dee_bosa and @GavinSBaker, getting the pregame started. It’s cold outside, but things are about to heat up. 🔥 https://t.co/rReacTNPft2d ago2.6K2713031.31%
$NVDA’s $150B buyback authorization is bigger than the market cap of around 84% of all S&P500 companies https://t.co/uLJ5qWldJDfrozen24d ago2.6K2503001.08%
More tokens for your $200? Claude wins in SemiAnalysis’s test. More useful work? That’s a different question. API list prices aren’t a productivity score. At BEP Research, we focus on cost per completed task. The workload and quality bar matter. Our breakdown below. https://t.co/R7ChHtPATgfrozen17d ago2.6K1101140.66%
Spot on, @ramahluwalia. I want AI that anticipates what I need and takes care of it. That’s the experience worth paying for. Also love the bank earnings lens: spending up, lending up, credit holding up. A healthier economy than the headlines suggest. https://t.co/tVOT7NKbA1frozen4d ago2.5K813010.51%
Can’t wait to dig into this treasure trove of data. Congratsfrozen22d ago2.5K701030.43%
Sign up your kids for @TrumpAccounts Everyone should participate and learn about capitalism and compounding at a young age Way to go @altcap 👶🇺🇸🚀📈💪 https://t.co/4vZOFChZu9frozen18d ago2.5K910010.43%
Zuck's yardstick for an AI lab: how big is your largest single supercomputer. Not total GPUs. One coherent cluster you can point at a training run. Prometheus in Ohio is scaling past 1GW. Hyperion in Louisiana goes to 5GW. Planned back in the Llama era, coming online now. Scale of the biggest coherent cluster is the moat. $META $NVDAfrozen9d ago2.5K1952011.07%
Sometimes the most entertaining outcome is the most likely 🤣 Last November, I wrote that Russia had the nuclear power and STEM talent to be an AI powerhouse. What war and sanctions cut off was access to chips, capital and collaboration. Now Putin’s envoy says US-Russia talks include joint data-center deals. Talks aren’t a deal. But the irony is hard to miss: the infrastructure of intelligence could become an incentive to cooperate. Power alone doesn’t make you an AI superpower. You need access to the rest of the stack anon.frozen5d ago2.5K704000.44%
I got a chance to actually play with Snacpchat specs The hardware and UI and UX is pretty impressive. Not heavy on your face like Vision Pro There is something here but it comes down to adoption https://t.co/OdTBe3bCezfrozen18d ago2.5K1712110.89%
You can see it today optics vs memory Ie $lite vs $MU Move the data faster Always been a memory bandwidth constraintfrozen21d ago2.5K800010.36%
I was in the room with Jensen: “Working capital will come. It gets solved.” Meta reports roughly $825B in future lease and contract commitments, many tied to AI infrastructure. Nebius just secured a $3.5B facility from Citi. Everyone is staring at the bill. We’re looking at what makes it so hard to stop spending. Every company is becoming an AI company. Someone has to build and finance the compute behind that transition. For hyperscalers, falling behind threatens the businesses they already have. Shareholders want more cash flow. Management sees competitors accelerating. Nobody wants to blink first. That changes the economics of growth. More compute. More power. More cooling. More capital committed before the revenue arrives. Jensen says demand will attract financing. Nebius shows it can. The next question is who earns the return on that capital. We’re watching the companies that can turn these enormous spending commitments into durable cash flow. Asset-light is getting heavy. One company’s capital bill is another company’s revenue. $NVDA $CRWV $META $NBIS5h ago2.4K2433031.35%
Everyone is focused on which AI has the best model. $META also has years of context on your interests and connections. Put that behind Muse and news gets a lot more personal. Knowing what matters to you is a powerful advantage. The social graph matters.frozen17d ago2.4K1112100.62%
Right @DavidSacks ? https://t.co/WfQQZwZo2Vfrozen24d ago2.4K700000.29%
Save money with Bloom Energy Follow the visual animation see if it makes sense We don’t have time to make turbine blades Time is money $BEfrozen27d ago2.4K801010.41%
NVIDIA just added $150B to the buyback authorization. $235B remaining through fiscal 2028. In June I called the old $119B tank about one year of fuel. Less than four months later, they refilled it. https://t.co/7WYIajrtMFfrozen24d ago2.4K1910030.96%
Excited to be joining my friends from Europe and team from Nvidia in Berlin 🍻🤖💚 Lots of exciting announcements coming $NVDAfrozen17d ago2.4K1912010.97%
NVIDIA's bear case just became a guest list lol. Miguel Burry's arc: Nov 10, 2025, depreciation. Nov 19, circular financing. Last night, antitrust and treble damages, off a report that unnamed neoclouds didn't get into GTC Berlin. That treats GTC like a birthday party where everyone deserves an invite. Treble damages go to a rival that sues and wins. No US conduct case has been filed. Meanwhile NVIDIA data center revenue went from $51.2B to $89.0B a quarter. Wouldn't you back your own ecosystem anon? Let’s have healthy bearish debatesfrozen5d ago2.4K1302020.72%
I accelerated my family’s Needoh hunting adventures with @bot and @usebland thanks to @mattyp video. It really works. Amazing. We are living in an agentic era https://t.co/5Ui8ti6Klpfrozen29d ago2.3K712010.48%
I remember when I was telling @FrancisSuarez that Miami needed an educational technical bench of students and researchers. Congrats to everyone Code 305!🚨frozen22d ago2.2K1005000.67%
6pm standing room only for Corweaves $CRWV Richard Ahlfeld at @NVIDIAGTC Berlin How do you architect for robotics? https://t.co/mj4hXuEXSF1d ago2.2K1120030.72%
Cool seeing the vision for recursively improving agents at DigitalOcean’s Open Intelligence Summit in SF today. Great connecting with @paddix , DigitalOcean’s CEO, and NVIDIA’s @adelhallak and @Shashank__Verma Run, evaluate, improve, repeat. Every loop needs compute, memory and verification. More soon on BEP Research $DOCN $NVDAfrozen9d ago2.2K2021001.04%
While the industry races for turbine capacity, Bloom already makes DC power. I'm still long. But what can the customer stop buying—and what comes back in the power bill? I interviewed Bloom's leaders. We stress-tested the economics. https://t.co/APdRjpKgjzfrozen28d ago2.2K1102030.72%
Zuck's whole strategy in one clip: own the stack, don't rent it. Meta's GPUs keep working after they're fully depreciated, and an H100 is worth more today than a year ago because compute is still constrained. Meanwhile app-layer agents like Instinct raise $1B at a $10B valuation, and a lot of that check goes straight to renting someone else's compute. When Meta ships a free agent to 3B users on clusters it already owns, the renters get crushed. $META $NVDAfrozen9d ago2.2K900250.73%
$2,000 iPhone. Some buyers shown a January delivery date. These iPhone Duo screenshots show Nov 20–30 for 256GB and Dec 8–15 or Jan 29 for 512GB, depending on location. The $AAPL question: strong demand, tight supply, or both? Long waits alone don’t settle it. https://t.co/sLxjlsNhNrfrozen6d ago2.1K402010.33%
Yo @thsottiaux with the reset, hope it hits! Codex https://t.co/xFQYiE21FBfrozen23d ago2.1K601000.34%
I'm gonna tell my kids this is AGI https://t.co/fB6cS70f3723d ago2.0K4013012.20%
AI is inflationary to build and deflationary to use. The spend hits first, the productivity shows up later. That’s the paradox behind the optimism.frozen19d ago2.0K1106030.98%
Cable cars in 1873. Robotaxis today. More than 150 years of progress, side by side on a San Francisco street. Good thing nobody decided the cable car was enough and we should “pace the frontier.” https://t.co/48DO7MYXfdfrozen9d ago2.0K1221010.80%
Watermelon is a big fruit 🍉 $META Scaling laws are holding. Access to coherent compute at scale wins. Anthropic reportedly asked Meta for compute. Meta declined, for now. Now Wonder Full lands near the frontier. But Meta also owns the consumer distribution. WhatsApp is already where people and small businesses communicate. Agents can meet them there. We wrote about this in April: AI for the rest of the world. Now add stronger models to that distribution advantage. The price war is about more than intelligence. It’s compute, distribution and who gets paid when agents finish the job. https://t.co/Lw2kaTT9rifrozen10d ago2.0K1210010.70%
Looks like the BEP Research interns were working hard and playing hard. We used OpenAI’s Astra, Blender and Higgsfield to turn @insane_analyst’s piece into a video for the visual learners. This took me back to my LED lighting days. Every light could work and still miss the spec.💡frozen18d ago2.0K1011060.91%
Ali’s measuring something that matters: did the agent actually finish the job, and can you trust the result? This connects directly to our BEP Verification Gap work. Add retries, compute and human oversight, and the metric gets even more interesting: cost per correctly completed task. Tomorrow at 8 AM PT: Part 1 of our new CPU series, on the computers behind the agents. Free to read.frozen11d ago2.0K902030.71%
On Monetary Matters Sept. 30, I drew on my LED company experience to explain the commodity cycle: “once supply catches up, watch out below.” Toshiba says its Philippine HDD operation aims to nearly double annual production capacity by FY2027 vs FY2025, measured in storage capacity. Our test: can demand keep pace?frozen20d ago1.9K907131.03%
I called NVIDIA the Fifth Avenue of compute in my Monetary Matters interview. Jensen’s point here gets at why: the same infrastructure can move between training, inference, models and customers. The tenants change. The address keeps earning. That flexibility matters when underwriting the useful life of $NVDA systems.frozen7d ago1.9K1730051.30%
AI agents can change company data or expose confidential information. Their permissions need enforcement outside the agent’s reach. NVIDIA’s OpenShell does that on CPUs. Optional Sentry adds an independent watchdog on BlueField-4 DPUs. OpenShell runs on x86 and Arm. Our $NVDA read: free software could create demand for extra silicon. Production adoption is the test. https://t.co/L1ZTgjwlvYfrozen24d ago1.9K1331000.89%
Great to see $META Meta and $AAPL Apple working together to push Local, open source AI together Big win for America https://t.co/fCmnCVm344frozen7d ago1.9K1411031.01%
Berlin looks good in NVIDIA green 🟢 Who’s headed to @NVIDIAGTC? I’ll be there with BEP Research, connecting the dots across AI, compute and the infrastructure behind it all. If you’re going, DM me. Let’s meet up 🇩🇪 https://t.co/RIBXKfp9pPfrozen11d ago1.8K1602011.03%
The AI revenue race flipped this year. It may have flipped back. Anthropic went from ~$9B run-rate at year-end to $65B by July and passed OpenAI in the spring. Axios says OpenAI is now near $70B. Two catches. Anthropic books cloud resale revenue gross, OpenAI books Microsoft's share net. And GPT-6.1 Sol now beats Opus 5.5 on Terminal-Bench at ~1/13th the cost per task. OpenAI is raising $30B+ at ~$1.4T. Anthropic's IPO could top $2T. Next scoreboard: the S-1.frozen23d ago1.8K801020.61%
CUDA haters, sit down. I was in the room at GTC Berlin when Jensen was asked whether CUDA’s moat is eroding. Jensen’s answer gets at the part you can’t tape out: an ecosystem built through 25 years of commitment to one architecture. His point was simple. CUDA alongside Arm. Alongside x86. Across clouds and industries. If you’re a researcher or founder building something new, you want your work to run where the infrastructure already is. That reach attracts developers. Developers add software. More software gives the next developer another reason to choose CUDA. That’s the flywheel. Sure, build a competitive chip. Then build all the reasons people choose to use it. You’re counting chips. NVIDIA’s compounding an ecosystem. $NVDA4h ago1.8K17001151.83%
I’m used to people handing out Bibles on the street. Outside SEMICON West in San Francisco, it’s “Principles of Building AI Agents.” Have you accepted automation into your life? https://t.co/5NHoXdEwiafrozen8d ago1.8K1800001.00%
Heading up to SF to see @paddix , @NVIDIAAI , and team make some agentic waves 🌊 Tune infrozen9d ago1.8K1102000.73%
After DigitalOcean's Open Intelligence Summit, I asked its team what 1,000 successfully completed agent tasks cost, with failures, retries, evaluation and tool calls included. I didn't get a worked bill. Part 2 of our CPU series builds a hypothetical one and uses it to test where the next dollar belongs: a better model, more memory, more CPU or better connections. Our view on NVIDIA's ($NVDA) Vera stands, and the bill defines which jobs it has to win. https://t.co/nxtom3cvnqfrozen5d ago1.8K2010001.18%
I didn't expect the best part of Nvidia's glass story to be the slab under the chip. The GPU and its HBM sit on it. In January I wrote that silicon warps past about 2,700mm². Glass is a shot at a bigger floor for more HBM. SCHMID is working with suppliers in Nvidia's chain. Wiring through the glass is still the hard part. Qualification isn't done. @jukan05 thinks Nvidia wants to make CoPoS work "one way or another." Curious who qualifies first.frozen26d ago1.8K711020.63%
2,000 sandboxes. One CPU. Agents need somewhere to get their hands dirty 🛠️ We wrote about Vera and the agentic CPU opportunity at BEP in July. Now @ivanburazin and @daytonaio are putting numbers behind it: 12.7 jobs/sec for Vera vs 10.9 for AMD EPYC 9755 in their test. Tools, code, tests, retries. Every step hits the bill. We’re digging deeper into this in our upcoming BEP CPU series. Love what you’re building, @ivanburazin 👏 You and the @daytonaio team are making this real. We had some fun visualizing it.frozen13d ago1.7K930020.80%
In February my feed was all AI video. I wrote that world models are memory monsters, and silicon built to train language models is a mismatch for them. In July I sat in AMD's keynote and wrote it was missing a layer: silicon, interconnect, software, no models. Today it agreed to buy Fei-Fei Li's World Labs for $8.2B in stock. Her team builds models that simulate 3D worlds. Chips take years, so a roadmap is a guess about the workload. Curious how much better the guess gets. $AMD $NVDAfrozen24d ago1.7K720000.53%
Weekends are for philosophy. Today’s question: why does everyone want more jobs until someone tries to build something? Data centers are creating jobs for electricians, fiber techs and construction crews. Let people build. https://t.co/Q8F5151oW3frozen19d ago1.7K832010.82%
Banks like $JPM are getting hit on the fear that consumer agents like Muse will massively accelerate chargebacks, fee disputes and penalty reversals. I think that fear is overdone. There will be guardrails around what agents can and cannot approve autonomously. Higher-risk disputes will get escalated to a human. Agents will make consumers more powerful. They won’t get an unlimited license to claw money back.29d ago1.7K711000.54%
Big week ahead in SF and San Jose ✈️ DigitalOcean’s Open Intelligence Summit, SEMICON West and dinner, then OCP. AI agents, chips and the infrastructure behind it all. Looking forward to getting out from behind the screen. Around next week? DM me. Let’s meet up 🤝frozen14d ago1.6K1202000.86%
Gassed up Codex with the Pro 500, no more resets and time outs please @thsottiaux lets gooo! https://t.co/c3JsE09D1gfrozen23d ago1.6K501000.37%
Bloom is up 11% since we got in at $252. Yesterday morning we told our paid subscriber chat: "Bloom dropped today because of that Oracle news, and we're buying on this dip." Oracle's force majeure notice to the Project Jupiter developer covers payments if the site misses 2028. Oracle says it's on schedule. Bloom says the 2.4 GW contract stands. The actual issue is gas. The line crosses half a mile of state land, and the land commissioner has said no twice. Paid subscribers heard it first. $BEfrozen27d ago1.6K802010.68%
RunPod reported $120M+ ARR in January. On Oct 8, our tracker captured six Secure Cloud GPU list rates rising 12.6%–17.7%. B200: $6.79 → $7.99/GPU-hr. One provider, not a market-wide index. Free brief + subscribe for our research: https://t.co/yDYpFTFf5N https://t.co/3ZPw5eYaF8frozen10d ago1.6K620020.62%
Funny anecdote I’m in Berlin for GTC my OpenAI Dot shows up once and while but isn’t available in the EU but @bot is saving my ass from @KLM losing my luggage to prepping me for my next days events1d ago1.6K400000.25%
Flagged Weaver in March. Now $CRDO has a working AMD FPGA implementation. The memory wall is where FLOPs go to wait. More capacity and bandwidth can matter as much as more compute. Next test: production silicon and real inference economics. $AMD https://t.co/9dDFNtGFopfrozen9d ago1.6K630040.82%
Great panel with @veelarco @EricSimons @spenserskates @ivanburazin. A directionally correct discussion on recursively self-improving agents doing load-bearing tasks. My buzzword bingo card is complete. 🤣 https://t.co/07iiVcFcQ8frozen9d ago1.6K600020.51%
NVIDIA-backed Iambic ($IAM) went public this week. Its own S-1 says about 8 cents of its 2025 R&D dollar went to compute, 31 to discovery and the lab, and 28.5 to the clinic. Two months into our AI-Science Basket: +20.5%, against -0.2% for the S&P 500 and -4.8% for biotech. The names that turn experiments into data rose 33.5%, or 16.8% without the Moderna readout day. Bioprocess and reagents rose 9.0%. The market bought the data. https://t.co/E1ZrPMaoJQfrozen7d ago1.6K700010.52%
FLYING CARSfrozen7d ago1.6K1000000.64%
“The most entertaining outcome is the most likely.” — @elonmusk 😂frozen14d ago1.5K900000.58%