New Ask Claude or ChatGPT about any stock, using Bullpen. Bullpen in Claude or ChatGPT. 7 days free
Bullpen.News From Edwin Dorsey
Get Daily Recap Feedback

NASDAQ

Nvidia CORP

NVDA

251 stories

$4.6T Market cap · 2026-11-06

Get NVDA news in Claude, ChatGPT or your own code. See plans

@aleabitoreddit

Serenity on X

Everyone’s a long term investor for 2027, 2028, or 2029 onward Until their portfolio experiences volatility months in. I still think: - $RKLB could end up the next $SPCX - $NBIS has the potential to be the next hyperscaler - $SIVE has moonshot potential to be the next $LITE - 1-3x P/E memory companies and $AXTI substrate bottlenecks are more durable than markets think And that it’s stupid to sell single digit fwd P/E AI companies or promising longs. And rotate that into 5% growth towel resellers at 47x P/E. Or 180x cyber companies, just because they’re going up. I’ve seen $AMD go from the $3 valley of doom to $400 as people thought it could be the next $INTC / $NVDA. And been through the Bitcoin $1000 -> $18,000 -> $3000 -> $100,000 cycle. Not advice on how to position or trade, but anyone on extreme leverage during those times got wiped out. While the underlying assets sharply recovered later. I do think AI, Space, and Physical AI are the future. So, I’m remaining long, and using this time to maximize exposure to the winners
@convequity 1 click

Convequity on X

$NVDA, $CBRS, and Huawei's Kirin 9050 Pro are three different bets on the same turning point: what the chip industry does when shrinking transistors stops paying. In AI chips, most energy goes into moving data, not into the calculation itself. In our view, as transistors shrink, calculations get cheaper much faster than the wires around them, so the industry's centre of gravity is shifting from ever-smaller transistors to shorter paths for data. That shift is why Huawei's logic folding — designing a circuit vertically to cut the distance data travels — looks potentially transformative and disruptive for the semiconductor industry to us. Here is the comparison in one line: one 16-bit floating-point calculation is the baseline, 1×, about 0.52 picojoules in NVIDIA's published GH200 energy model. Every figure below is the energy spent moving data, measured against that one calculation. 1. Moving data to local SRAM next to the computing units: Cerebras uses about 2× the baseline. NVIDIA's highly optimized local-memory path uses about 5–15× the baseline. 2. Moving data across the chip: we estimate Cerebras's core-to-core network at about 12–37× the baseline. NVIDIA's optimized path through L2, its shared on-chip cache, is about 60–90× the baseline. 3. Moving data off the chip to external memory: Cerebras's MemoryX path is about 330–800× the baseline. NVIDIA's optimized path to HBM, its external high-bandwidth memory, is about 230–290× the baseline. The energy cost is set by how far the data travels: a short hop costs single-digit multiples of one calculation, a trip across the chip costs tens of times, and a trip to external memory costs hundreds of times. If moving data is the real constraint, then shortening paths, which is what Huawei is doing with logic folding, matters more than another process node — for NVIDIA and Cerebras as much as for Huawei. These are engineering estimates, workload-dependent, not measurements of complete systems. The optimized block transfers behind NVIDIA's figures are already counted above; what remains is the physical cost of the wires and memory themselves. In our view, Cerebras's commercial case depends on whether its compiler can keep data close to computation across enough workloads. NVIDIA's advantage is a broader range of software, not a way around the energy cost of moving data.
@DeepSailCapital 1 click

Deep Sail Capital on X

Just a bit of my thoughts on the AI datacenter buildout: Texas has already stopped new data centers back in August, and that's the most important fact the market missed today. 1/ Abbott paused it in August. ERCOT stopped energizing new data centers of 75 MW or more and shelved a ~200 GW study batch. TCEQ stopped issuing permits. Hinojosa's moratorium will keep the existing pause in place on New permits instead of lifting it. She also said it targets new permits, not projects already under construction. So that just means next project goes to a friendly locale, not Texas. 2/ Ohio was already moving. DeWine froze the data center sales-tax exemption for new projects back in June. Acton's conditional moratorium is a pay your own power costs, use union labor, build on brownfields. That's a cost hit, not a ban outright. 3/ Congress can't ban anything. The Sanders/AOC moratorium bill would face a Trump veto through Jan 2029. 4/ What's already built or nearly built keeps going: • Stargate Abilene: 4 of 8 buildings live, the rest due this month • Fairwater Atlanta: ~636 MW running • Rainier Indiana: ~1 GW running • Hyperion Louisiana: not in play at all About 56.7 GW is under construction nationally and largely outside the pledges. 5/ Demand didn't change this week: • $ANET Q3 revenue $3.66b, +58.5% y/y, $360m above guide • $CLS guiding 2027 to $40b, roughly double 2026 • $MOD data center sales +95%, record backlog, a $4b+ deal for 2027–29 • Hyperscalers are on pace for $803b–$828b of 2026 capex and they're guiding 2027 higher 6/ The market half-figured it out. Texas-heavy builders and power names fell 7–23% ($HUT -23%, $VST -15%). Suppliers with national backlogs barely moved ($ANET -0.4%, $NVDA -0.8%). $NUAI is a zero now for sure. 7/ Where it's real: new 2027–28 capacity in TX/OH/GA gets slower and pricier, so it will shift to different locations. New projects go to friendly locales, TX/OH/GA lose out. 8/ AI component suppliers are still maxed out for the next few years, that demand doesn’t change, just where they ship it to does. 9/ Neoclouds still suck. Ain’t nothing gonna change that.
@DeepSailCapital 1 click

Deep Sail Capital on X

Just a bit of my thoughts on AI: Texas has already stopped new data centers back in August, and that's the most important fact the market missed today. 1/ Abbott paused it in August. ERCOT stopped energizing new data centers of 75 MW or more and shelved a ~200 GW study batch. TCEQ stopped issuing permits. Hinojosa's moratorium will keep the existing pause in place on New permits instead of lifting it. She also said it targets new permits, not projects already under construction. So that just means next project goes to a friendly locale, not Texas. 2/ Ohio was already moving. DeWine froze the data center sales-tax exemption for new projects back in June. Acton's conditional moratorium is a pay your own power costs, use union labor, build on brownfields. That's a cost hit, not a ban outright. 3/ Congress can't ban anything. The Sanders/AOC moratorium bill would face a Trump veto through Jan 2029. 4/ What's already built or nearly built keeps going: • Stargate Abilene: 4 of 8 buildings live, the rest due this month • Fairwater Atlanta: ~636 MW running • Rainier Indiana: ~1 GW running • Hyperion Louisiana: not in play at all About 56.7 GW is under construction nationally and largely outside the pledges. 5/ Demand didn't change this week: • $ANET Q3 revenue $3.66b, +58.5% y/y, $360m above guide • $CLS guiding 2027 to $40b, roughly double 2026 • $MOD data center sales +95%, record backlog, a $4b+ deal for 2027–29 • Hyperscalers are on pace for $803b–$828b of 2026 capex and they're guiding 2027 higher 6/ The market half-figured it out. Texas-heavy builders and power names fell 7–23% ($HUT -23%, $VST -15%). Suppliers with national backlogs barely moved ($ANET -0.4%, $NVDA -0.8%). $NUAI is a zero now for sure. 7/ Where it's real: new 2027–28 capacity in TX/OH/GA gets slower and pricier, so it will shift to different locations. New projects go to friendly locales, TX/OH/GA lose out. 8/ AI component suppliers are still maxed out for the next few years, that demand doesn’t change, just where they ship it to does. 9/ Neoclouds still suck. Ain’t nothing gonna change that.
@DeepSailCapital

Deep Sail Capital on X

Just a bit of my thoughts on AI: If AI buildout is really materially slowed down it’s a death sentence to American Economy / our AI pole position in the world. AI datacenter buildout saved us from a recession a few years ago, it is the only thing holding us back from one right now. An extremely bad outcome is China takes the lead in AI, US businesses use said Chinese AI to cut US jobs and the money saved is shipped to China, so we are net negative as a country. Now on to why I don’t think AI build cadence will be materially changed and it’s all posturing / the easy trade for tourists. Texas has already stopped new data centers back in August, and that's the most important fact the market missed today. 1/ Abbott paused it in August. ERCOT stopped energizing new data centers of 75 MW or more and shelved a ~200 GW study batch. TCEQ stopped issuing permits. Hinojosa's moratorium will keep the existing pause in place on New permits instead of lifting it. She also said it targets new permits, not projects already under construction. So that just means next project goes to a friendly locale, not Texas. 2/ Ohio was already moving. DeWine froze the data center sales-tax exemption for new projects back in June. Acton's conditional moratorium is a pay your own power costs, use union labor, build on brownfields. That's a cost hit, not a ban outright. 3/ Congress can't ban anything. The Sanders/AOC moratorium bill would face a Trump veto through Jan 2029. 4/ What's already built or nearly built keeps going: • Stargate Abilene: 4 of 8 buildings live, the rest due this month • Fairwater Atlanta: ~636 MW running • Rainier Indiana: ~1 GW running • Hyperion Louisiana: not in play at all About 56.7 GW is under construction nationally and largely outside the pledges. 5/ Demand didn't change this week: • $ANET Q3 revenue $3.66b, +58.5% y/y, $360m above guide • $CLS guiding 2027 to $40b, roughly double 2026 • $MOD data center sales +95%, record backlog, a $4b+ deal for 2027–29 • Hyperscalers are on pace for $803b–$828b of 2026 capex and they're guiding 2027 higher 6/ The market half-figured it out. Texas-heavy builders and power names fell 7–23% ($HUT -23%, $VST -15%). Suppliers with national backlogs barely moved ($ANET -0.4%, $NVDA -0.8%). $NUAI is a zero now for sure. 7/ Where it's real: new 2027–28 capacity in TX/OH/GA gets slower and pricier, so it will shift to different locations. New projects go to friendly locales, TX/OH/GA lose out. 8/ AI component suppliers are still maxed out for the next few years, that demand doesn’t change, just where they ship it to does. 9/ Neoclouds still suck. Ain’t nothing gonna change that.
@aleabitoreddit

Serenity on X

For companies like $SIVE / $AAOI, the demand imbalance for lasers are so severe That even a 20-30% overall demand reduction from a hypothetical DC slowdown would likely not eliminate the shortage. We're seeing this before the next architectural shifts, where $NVDA Jensen Huang expects 15x optical volume for CPO scale-up. And as more hyperscalers commit to NPO as seen with $LITE new $5B UHP laser contracts. I'd argue this is more nuanced since I can't say the same for finished 800G/1.6T pluggables or lower end 70mW-100mW laser ranges that are currently bottlenecks. But this argument is more toward higher spec lasers. $AAOI is vertically integrated and has 300-400 mW capabilities, so they have the capability to shift attention to merchant lasers over supplying end-to-end optical transceivers. $SIVE is more pure play on this laser bottleneck, but they have the capabilities for 200mW + higher margin multiwavelength NPO/CPO lasers. I'd argue this is specifically the case for bottlenecks where demand imbalances are so deep (creating higher long term visibility, such as HBM). Whether it's large body substrates from $ATS that's bottlenecking Marvell + $AMD or lasers from $LITE that are sold out into 2029. or liquid transformers that have 4-5 year lead times. They they would still sell out of capacity even if overall demand decreased. If markets de-risked, it would make sense to hit optical module manufacturing players, maybe server rack assembly, to companies that still require heavy financing like Neoclouds. And obviously haircut growth valuations (but not to the point where demand collapses to 0). Yet we're seeing everything in AI trade is getting lumped up together as something that requires heavy financing + not profitable.
@aleabitoreddit 1 click

Serenity on X

If a company is valued at $2B, has ~$1B in cash/inventory... And makes $354M FCF a quarter. Then returns 30-50% of FCF for dividends next year. I'd argue it's more of a matter of time before shareholder returns/fundamentals, (especially memory) overpower forced selling. But stock prices can look really dislocated from actual performance for periods of time, like now. Same with photonics, but that's more of a 2027-2028 growth type story with CPO + more optical content/xpu. Laser companies like $LITE, $AAOI, and potentially $SIVE depending on quals, would likely be printing few years after capacity expansion comes online. And it's already been largely funded... Lumentum with $NVDA + hyperscalers, AOI with $1.6B+ in ATMs, Sivers with 700M SEK from institutions. Compared to other sectors that still require a lot of capital for their buildouts.
@convequity

Convequity on X

Larger AI chip packages are pushing manufacturers toward glass substrates. We regard LPKF Laser & Electronics ($LPKFF) as one of very few suppliers that can make the tiny, closely spaced holes through the glass reliably and in volume. 1. The silicon chips and the organic (resin-based) substrate beneath them expand by different amounts as the package heats and cools. A larger package makes that mismatch harder to manage, raising the risk of warping and damaged connections. Glass expands more like silicon. 2. $NVDA is requiring its Taiwanese and Korean suppliers to deliver glass substrates by 2028. 3. China's BOE is reportedly meeting with Samsung on a glass substrate partnership. 4. Producing these holes in volume is difficult. The holes, called through-glass vias, must be made without cracking the glass and then filled with copper. Our view is that LPKF's process, a laser pulse followed by a chemical etch, is built for that hole-making step. We are watching for its customers to move from development into production.
@aleabitoreddit 1 click

Serenity on X

The rhetoric on X and from Elon leans overwhelmingly apocalyptic on Democrats winning. It's a mini bubble here, but I think people forget AI is bipartisan national security priority, and that $NVDA's historic rally + AI buildout acceleration happened under the Biden admin. Anyone can extrapolate a politician's negative view on AI to doompost indexes or that " $CRWV is going to go down". But in all reality, even with a Democratic governor, we'll likely just see some headwinds to individual developers, and incentives like "clean-energy tax credits" moving to other areas in AI like solar manufacturing. Rather than any market-wide doomsday scenario that the opposing party likes to portray. Even Bitcoin/Ethereum performed better under the negatively perceived regulation of Gary Gensler. Just like any major event, everyone looks at the worst case scenario like helium disruption shutting down Samsung, $TSM, and the memory trade... Then markets move on few weeks later.
@convequity

Convequity on X

Compute & Accelerators screen (43 names). Our Rule of X = forward growth x a sector coefficient + free cash flow margin - SBC %. The coefficient starts from Bessemer’s 2.3 for software and is scaled by how much profit a sector earns per $ of growth versus its cost of capital. More profit and cheaper capital means growth counts for more. Compute mostly runs 2.3 to 2.7. Colours: green = high score, cheap, or price down (more upside from here). High score + attractive valuation + flat or falling price: $NVDA (DCF 91% discount), $AVGO (94%), $ASML (EV/FCF 24.8x). High score but priced for it: $AEHR (EV/GP 95x), $AMD (EV/FCF 125x). Data from Koyfin, prices at 30 Oct close. A screen, not advice. Do your own research.
@aleabitoreddit

Serenity on X

Agility Robotics merger is now approved! $AGLT is expected to begin trading November 5th, following the November 4 closing... As the first US pure-play humanoid company! The company is backed by $NVDA, $AMZN, Foxconn, Softbank, and others... to automate human work in warehouses and manufacturing. ~61.48% of the 41.4M public shares were redeemed (~25.45M shares), leaving ~15.95M public shares tradeable. Given heavily reduced share count, as a warning, the stock may experience significant volatility. Disclosure: Not a recommendation, purchased $CCXI shares today after seeing the news and currently hold a position.

Nvidia: Trade The Fear With NVDL

Nvidia guides for 70% growth despite supply constraints and near-100% demand, while NVDL is framed as a tactical trade amid credit, margin…

NVDA $4.6T
Benjamin C. Barry
@aleabitoreddit 1 click

Serenity on X

$LITE standalone would likely be up on earnings after their new $5B CPO/NPO announcement (separate from $NVDA). The margins on that are likely 60%+… which shows how profitable laser players would be at scale. (Greensboro alone would be $5B revenue/year) In terms of the $MRVL deal: - Extremely positive for Marvell strategically. Also makes remaining laser players very valuable - $LITE shareholders can cry though about the 1+ year overhang until deal goes through in 2028. There’s less independent price discovery as well since a lot of it will be pegged around 4.5 Marvell shares : 1 - I think Lumentum shareholders probably would be happier if Marvell shares didn’t crash before the deal was announced.. since it was originally valuing $LITE at $115B with a 23.9% premium. - Marvell dropping does show the downside of the deal straight away though, if it rose then $LITE shareholders might be a tad happier given valuation. Having a fixed dollar MC value floor probably would have been better than the 4.5x:1 ratio. There was a lot of long $LITE, short ___ pressuring other photonics valuations, so at least those trade pairs might unwind now
@aleabitoreddit

Serenity on X

Following the Lumentum / Marvell deal: My opinion is that independent Western laser players, like $AAOI, $COHR, and $SIVE became a lot more strategically valuable. $MTSI and $SMTC should get some attention too as emerging laser players in 2027. If I had to guess: 1. Companies competing against Marvell, or not in the $NVDA ecosystem bandwagon, such as $AMD, are likely going to rethink their relationship with $LITE. If getting bottlenecked by Nvidia capacity agreements wasn't enough... Incentives around qualifying + bringing up other laser players, seems very necessary now. Since you generally don't want competitors like Marvell to have influence or visibility over your roadmaps, IP, capacity, pricing, etc. even if there is a "firewall". 2. $MRVL acquiring Lumentum opened pandoras box in a way of acquiring a critical chokepoint all your competitors + buyers use. $LITE was used by your major hyperscalers and across AI supply chains. Now, your leaders might wonder if weaponizing capacity agreements might not be as effective.... If someone else can just buy the entire company. There's only a few major public merchant Western players out there, and I'm also sure many others will want to follow the Marvell / Broadcom route via vertical integration, given I think $MRVL saves around $1B+ in costs/year. So M&A premiums likely went up. $SMTC was actively looking to buy laser capacity via M&A so I'm sure they're looking at $SIVE given their 100M CW DFB laser expansion + tiny MC. $JBL also probably wants to vertically integrate down the road too so $SIVE is also an acquisition target for their 1.6T program. $AMD was looking for ELSFP + CW capacity given Nvidia bottlenecks, so I'm sure they've looked at $AAOI. can go on and on about other ASIC players (like in Taiwan) that likely are looking at vertical integration too. Since multiple players use a small subset of laser suppliers, if competitors just go ahead and buys them... There's a lot pressure to M&A for their own supply chain defensibility. TLDR: Independent laser suppliers should capture a material strategic revaluation since $MRVL + $LITE, sets a new precedent + implications. We'll see what happens.
@convequity

Convequity on X

Further thoughts on the $CBRS short-seller report I've already quote-tweeted this short-seller report, but I think it's quite funny. These days, a lot of tech short sellers seem to have a poor understanding of the technology itself. Perhaps they're better at timing market liquidity and identifying moments when sentiment is most vulnerable. Going back to the report's central claim, I think it's arguing that Cerebras hasn't adequately addressed the challenges of using a giant substrate to support an entire wafer, or of maintaining reliable connections across such a large chip. It's similar to the challenges facing $NVDA's Rubin Ultra. The idea is to connect four compute chips, with a combined size roughly eight times the reticle limit. At that scale, thermal expansion becomes a major issue, as does maintaining consistent communication between the chips. For Cerebras, these are certainly significant challenges. That's why it took so many years to turn the original idea into something that actually works. I would say the third-generation Wafer-Scale Engine, WSE-3, was the first generation to become commercially attractive. Earlier generations had lots of problems to work through, much like other chips, for example Google's TPU from v1 to v3. Even WSE-3 hasn't solved everything. One of the biggest remaining issues is parametric yield. The simplest way to explain it is that, when a chip spans an entire wafer, you can't guarantee that every part will operate reliably at the same high frequency. (Frequency is how many cycles per second the chip runs, so a higher frequency means more compute from the same silicon.) You therefore have to lower the operating frequency of the whole system. A normal chipmaker would cut the wafer into many separate dies and set each one's speed on its own. The argument is that those dies could then run at frequencies 50% higher, or perhaps even twice as high. Cerebras keeps the whole wafer as one chip instead, so it gives up some of the compute the same silicon could otherwise deliver. In return, it gets an enormous amount of compute and on-chip memory connected as one wafer-scale chip. Data moves between the cores over wiring on the wafer itself, rather than through slower, more power-hungry links between separate chips. That is what lets Cerebras serve AI models so quickly, especially for inference. That's the trade-off, and parametric yield has been a constraint for Cerebras. The workaround has been to lower the specifications and operating frequency, leaving enough headroom for the system to work reliably. But that also means leaving some of the wafer's potential unused, despite manufacturing it on an expensive, supply-constrained TSMC N5 process, and potentially on more advanced nodes in the future. That's why I find the latest CS-4 rack-level system, using the WSE-3T, interesting. My understanding is that WSE-3T uses the same wafer design as WSE-3 but delivers more than 50% higher performance. They've apparently found new ways to address the parametric-yield issue, allowing them to unlock frequency headroom they previously couldn't use reliably. It looks as though they're actively addressing the problem and making good progress. The earlier implementation may have been suboptimal, but even that version was already demonstrating commercial competitiveness. This is a common problem with short sellers today: they build a thesis around old information while the technology is evolving rapidly. They point to a technical limitation that may still exist, but fail to distinguish between a problem that makes a product unworkable and one that merely reduces its performance. Between roughly 2015 and 2021 or 2022, these issues may genuinely have been showstoppers. By late 2026, I don't think that's necessarily true anymore. They may still hold performance back somewhat, but they don't prevent the product from working. Meanwhile, Cerebras has the advantage of being the only company to have made a wafer-scale chip work. In my view, that benefit outweighs the remaining limitations. The short seller doesn't seem to recognize that distinction. I think this also explains why Oracle, AWS, OpenAI, and other companies have begun testing Cerebras chips this year. The technology seems to be reaching a point where these problems are largely, though not entirely, resolved. Previously, perhaps they were only halfway to a solution, and the product wasn't sufficiently practical. Now that they're much closer, large-scale commercial deployment becomes more attractive to customers. Of course, it remains to be seen how this plays out. They still haven't crossed the deployment threshold I have in mind, and once they do, we may see new problems emerge. For now, though, my sense is that these are well-known issues among people familiar with the technology. The short seller is largely repeating things that those people already understand. The more interesting question is the direction of progress: how quickly are the problems being resolved, and what does that mean for the product's competitiveness? The report doesn't really address that. The rest of the report seems to cover the usual short-seller material, raising fear, uncertainty, and doubt about adoption and other peripheral issues. But the central technical thesis strikes me as quite weak. It reminds me of previous short reports on companies such as $SMCI.
@AsifSuria

Asif Suria on X

An excerpt from our article about Expand Energy $EXE that we published this morning after Danny Brown took over as the new CEO this week: Natural gas may not be the first thing that comes to mind when you think about the AI boom. But as companies race to build data centers, finding enough electricity to power them is becoming just as important as getting hold of the top GPU chips from NVIDIA $NVDA Expand Energy is a company you might recognize by its old name, Chesapeake Energy. Chesapeake has quite a history. It was one of the biggest names in America’s shale gas boom, built aggressively by its co-founder Aubrey McClendon. But its rapid expansion came with enormous debt, and the company eventually ended up in bankruptcy in 2020. Today, it looks very different. Following its merger with Southwestern Energy, Chesapeake became Expand Energy, creating the largest natural gas producer in the United States. The company is performing well, with our quant model IA Score rating it a 90 (a higher score is better), even though natural gas prices haven’t enjoyed the same strength as oil. That is what caught my attention. Natural gas producers typically do well when gas prices rise and struggle when they fall. Yet Expand appears to be building a business that could benefit from more than just a recovery in commodity prices. And now, with Danny Brown taking over as CEO, the company is entering another interesting chapter. To understand the opportunity and why this company is worth watching, we need to rewind the tape and go back to where it all started: Chesapeake Energy.

Investor Letters

Letters that discuss NVDA.

Every letter that discusses NVDA

Tickers on this page

Companies the stories above are also about, besides NVDA.