META — expert X mentions

Mizuho Securities: CPUs & GPUs Market Forecasts & Growth > Shipment Growth: Industry server CPU shipments are forecasted to reach 35 million units in 2026 and grow to 50 million units by 2027, representing a 40% year-over-year increase. > Long-Term TAM: The long-term Total Addressable Market (TAM) estimate for 2030 has been raised to $170 billion (up from the previous $107 billion forecast), driven by higher CPU-to-GPU ratio assumptions for AI inference servers. > CPU-to-GPU Ratios: The CPU-to-GPU ratio on AI servers is accelerating and is expected to approach 1:1 by the end of 2027 or 2028. Supply Chain & Technical Bottlenecks > DRAM Constraints: A critical bottleneck exists in DDR5/LPDDR5 supply, with a projected fulfillment ratio of only 70% over the next 12–18 months. > Demand vs. Supply Gap: Based on current models, the 2027 demand for DDR5/LPDDR5X (over 300 billion 1Gb equivalents) significantly exceeds the projected supply (220–250 billion 1Gb equivalents). > Potential Risks: The shortage of key materials—DRAM, substrates, and passives—is expected to persist through 2027 and could pose downside risks to downstream server assemblers, potentially leading to lower server rack output. Key Player Insights (2027 Forecasts) > Nvidia: Expected to reach 5.0–6.0 million units for the Vera CPU, including 2.0–3.0 million units specifically for agentic AI stack racks. > Google: Axion CPU production is projected to increase more than 2x year-over-year, aligning with the growth trajectory of TPU units. > AMD: The N2 Venice CPU is forecasted to exceed 6.0 million units. GPUs/ASICs Market Growth Projections > Rapid Expansion: The total AI ASIC market is projected to grow from 4.1 million units in 2025 to 24.0 million units by 2028. > Volume Drivers: The growth is driven by substantial increases in deployment by major hyperscalers including Google, Amazon (Annapurna), Meta, Microsoft, and OpenAI. > External Demand: The market for external (non-Google) AI ASIC units is expected to surge from 0.6 million in 2025 to 7.2 million by 2028. Key Hyperscaler Activity > Google (TPU): Continues to be a dominant player, with total shipment units increasing from 2.5 million in 2025 to 7.1 million by 2028. > Anthropic: Significant ramp-up is forecasted for their "TPU Ironwood/Sunfish" chips, moving from 0.6 million units in 2026 to 6.2 million units by 2028. > Amazon/Annapurna: Shipments for the Trainium line are projected to double from 1.5 million in 2025 to 3.6 million by 2028. > Meta: Rapid scaling of MTIA chips is expected, growing from 0.1 million units in 2025 to 2.7 million units by 2028. Technical Trends > Advanced Packaging & Nodes: There is a heavy reliance on sophisticated packaging technologies like CoWoS-L and CoWoS-S, and advanced foundry nodes including N2, N3, N4, N5, and A16. > HBM Integration: Nearly all listed high-performance ASICs utilize High Bandwidth Memory (HBM), with a transition toward newer generations such as HBM3E and HBM4/4E to meet performance demands. > ASP Variance: Average Selling Prices (ASP) range significantly, from approximately $2,000 for entry-level models to as high as $40,000 for top-tier specialized chips like the TPUv10. $DRAM $EWY $MU $GOOGL $AMKR $TSM $ASE $NVDA $AMD $AVGO $MRVL $INTC $MSFT $META

$META in talks to rent up to $10B in AI compute to Anthropic over two years. GPU+HBM clusters are becoming enterprise revenue engines. $MU $SKHY $DRAM. Healthy Economics Incoming. Claude is the enterprise standard for complex, high value problems. Most expensive frontier model per task. Highest KV cache footprint. Highest HBM demand per token. Enterprise doesn't mean casual usage. It means contracted, always on workloads running around the clock. Every dollar Meta charges Anthropic for enterprise compute flows directly through GPUs and HBM.

$META 👀 Interesting move. A potential $10B Anthropic compute deal shows how serious Meta is about securing AI capacity. The next phase of AI won’t just be about models. It will be about chips, data centers, and computing power. #Investing #Stocks #WallStreet #NASDAQ #SP500 #NVDA

$META in talks to lease compute to Anthropic in a $10B dollar deal. Seems like they saw how profitable $SPCX $45B compute deal was. But just goes to validate Neocloud business models like $NBIS, $IREN, and co if hyperscalers are copying their homework. Source: NYT https://t.co/3Mu3UXQtFT

Follow the $META compute saga, because it's the best FUD case study of the year. All numbers, no vibes. July 1: Bloomberg reports Meta is exploring selling excess AI compute through "Meta Compute." The single word "excess" nukes the sector: Then look at what Meta actually did next. July 8: Starts production of its in-house AI chip, Iris, this September. Doubles compute target from 7GW to 14GW by 2027. July 14: Expands the Richland Parish, Louisiana data center to 5GW. Announced investment goes from $10B to $50B+. Bloomberg reports total site cost could top $250B including chips. $1B+ in local infrastructure, 1,000+ operational roles. Same week: Morgan Stanley raises hyperscaler capex estimates to $1.23 trillion for 2027 and $1.4 trillion for 2028, up 9-10%, with Meta alone at $250B in 2028. And now the best part. $META is reportedly in final talks with Anthropic on a ~$10 billion compute deal per. So the "excess compute" that flushed the sector was never a glut. It was inventory. Meta saw what compute rents for and realized its spare capacity is a product. The market sold "excess" as weakness. The data says it was a company discovering demand so strong it pays to become a landlord. The neoclouds got flushed on the exact news that validates their business model.

Meta $META in talks to lease computing power to Anthropic. Deal worth as much as $10B over 2 years. NYT: "Anthropic proposed the deal in June and Meta is considering it, said the people, who were not authorized to discuss confidential conversations. While the specifics were in flux, Anthropic would pay Meta in monthly increments over the two-year period, the people said. The companies would be able to opt out of any agreement early, they added. Anthropic’s proposal to Meta was about a third of the size of a deal that the A.I. start-up signed with Elon Musk’s SpaceX in May. Under that agreement, Anthropic is paying the rocket company $45 billion over three years — or $1.25 billion a month — for computing power. The deal included similar provisions that let either company exit the agreement early."

$META BREAKING: Meta is in talks to rent computing capacity to Anthropic in a deal worth up to $10B over 2 years, as per the New York Times. Meta went down today on news that they hired a senior AWS executive to help build their cloud division. Why? Because it was confirmation that they are doubling down on capex and selling compute, which means their FCF will take a hit. Ironically, the semis should have been up on this news but the Kimi K3 headlines have kept them down. Now, Meta is slightly reversing on this headline about selling compute to Anthropic…which begs the question…if the market is happy about Meta selling compute to a big lab, why are they upset about Meta spending more on capex in order to build that compute business? Tons of weird narratives conflicting with each other, but it looks like META is officially getting into the neocloud game.

$META 🚨 META IN TALKS TO LEASE COMPUTING POWER TO ANTHROPIC: NYT Deal worth over 10 BIllion Dollars for 2 years Sometimes they are leasing themselves, building data centers, now they leasing from others! Not sure what $META is doing

$META 🚀 Interesting move from Zuckerberg. Hiring senior AWS infrastructure talent suggests Meta is thinking bigger about AI, data centers, and possibly cloud services. The next phase of AI is not just about ideas. It’s about having the power and infrastructure to make them real.

I just published my Q2 channel check and alternative data report. 1. $AMZN, $MSFT, and $GOOGL data. One cloud provider has seen significant momentum this quarter. 2. $MSFT Copilot usage trends and headwinds facing SaaS. 3. Usage on ad tech platforms ( $META ) showing strong acceleration. https://t.co/QpomSOsHkP

You can’t make this up! Since June 25th: $MU -30% ish $META +30% ish Rotate, rotate, rotate! To perfection.

@michaelsikand DRAM 20%+ hike for next quarter, $SNDK LTAs with $META, $MU 16+ LTAs. Market: proceeds to sell off memory. 800G transceiver revision sharply upward? Lasers completely sold out into early 2029? Market: sells off photonics and laser companies.

@meikozz_2026 I don't see any fundamentally wrong. There's probably going to be large corrections from time to time flush out margin/leverage before things move higher. And this month seems like that time of year? $POET confirmed your big optical giants like $LITE, $COHR are completely sold out for the next 2 years, and likely into 2029 for photonics. Innolight confirmed 800g transceiver upward revisions 3 days ago, so that should be positive for $AAOI and the others. Samsung became the most profitable company in the world, and continues to project DRAM hikes for future quarters. $MU signed 16+ LTAs showing memory demand is structural... $META + hyperscaler capex plans are on the higher end of projections. I wouldn't conflate short term price movements with longer term trends. And as seen with $AEHR, recoveries tend to be extremely fast (eg. 1M of corrections wiped out overnight).

Morgan Stanley: AI/Data Center Debt Issuance > Massive Year-Over-Year Surge: Total global credit market debt issuance for AI and data centers is growing at a rapid pace. YTD 2026 issuance has already reached $336bn, easily eclipsing the $217bn total for the entirety of FY 2025. > Aggressive Projections: The full-year FY 2026 forecast stands at $580bn, representing a 167% increase compared to FY 2025. $GOOGL $AMZN $META $ORCL $MSFT

In Ackman's latest interview he divides the AI market, stating that it is not a bubble at the foundational layer, but highly risky at the software/model layer. He sees "near infinite demand" for compute. He believes infrastructure providers such as Microsoft, Amazon, Meta, and even SpaceX will earn massive returns on their data center investments. He is highly skeptical of pure model layer businesses like OpenAI and Anthropic. He notes they face massive capital burn rates and an existential threat from open-source models such as Chinese model and SpaceX that are driving the cost of AI access effectively to zero. A business model reliant on charging for proprietary models will collapse if open-source alternatives are essentially free. $AMZN $META $MSFT $SPCX

I still can’t believe $META news tanked the entire neocloud sector. Then once the sector sold off, they doubled its compute target from 7GW to 14GW, expanded its Louisiana data center from $10B to $50B+, put its own AI chip into production, and locked in multi-year memory supply deals. The “excess compute” was never a demand problem. It was a company with so much demand for its capacity that renting it out prints money.

A TON OF THINGS HAPPENED IN THE STOCK MARKET TODAY. Here's a full recap: 1. Trump says the U.S. is reinstating the blockade and will now charge a 20% fee on all cargo passing through the Strait of Hormuz, effectively putting a tariff on one of the world’s most important shipping routes. He also said the U.S. is “probably just going to take over the Strait,” arguing America has guarded it for 50 years without being paid, while warning Iran that the U.S. is prepared to hit them “very hard” tonight and tomorrow. 2. Apple $AAPL is expected to launch a base M6 this fall, but reportedly skip the M6 Pro, M6 Max, and M6 Ultra, moving directly to M7. The M7 Ultra is designed to support up to 1.5TB of memory, roughly double the planned M5 Ultra capacity, and could help power Apple’s future AI server strategy. Apple is also already developing M8 chips with greater AI capabilities, including 2028 chips moving to a 1.4nm process. $AAPL Apple hit an all-time high today. 3. TSMC $TSM reported record Q2 revenue driven by AI demand. June revenue came in around $13.8B, up 6.2% MoM and 67.9% YoY. Q2 revenue was about $39.6B, up 36% YoY and above the roughly $39.4B estimate. First-half 2026 revenue reached about $75.0B. 4. Meta $META is expanding its Richland Parish, Louisiana data center to 5GW of compute capacity, raising the announced investment to over $50B from the original $10B plan. Bloomberg reports the total site cost could exceed $250B when including chips, though Meta has only publicly disclosed $50B. Meta also plans more than $1B in local infrastructure improvements and says the site will support 1,000+ roles once operational. 5. Newly disclosed court records show Apple $AAPL is suing OpenAI, alleging misappropriation of trade secrets likely stemming to a new consumer device. The Information reports OpenAI’s first consumer AI device could launch as early as February, though Apple’s lawsuit could disrupt the timeline. OpenAI has reportedly explored several hardware concepts, including a display-less smart speaker, smart glasses, a digital voice recorder, and a wearable pin. 6. Penguin Solutions $PENG announced a proposed $650M private offering of convertible senior notes due 2031. Proceeds are expected to fund capped call transactions, refinance existing notes, and repay $100M under its credit agreement as part of a broader capital structure move. 7. The top 10 most active options today by contracts traded were $NVDA with 2.6M contracts, $TSLA with 2.0M contracts, $AAPL with 1.6M contracts, $AMZN with 676K contracts, $MSFT with 673K contracts, $MU with 609K contracts, $META with 560K contracts, $DRAM with 499K contracts, $INTC with 460K contracts, and $SPCX with 431K contracts. 8. Korea-focused leveraged ETF AUM peaked above $45B in June, but after the recent pullback has fallen to roughly $28B. Exposure remains heavily concentrated in a few key areas: $14B in SK Hynix-linked products, $6B in Samsung-linked products, and $8B in KOSPI 200-linked products. 9. Cantor Fitzgerald reiterated Rocket Lab $RKLB at Overweight with a $96 price target. Analyst Andres Sheppard highlighted Rocket Lab’s successful U.S. Space Force VICTUS HAZE mission, marking the first time a prime contractor completed an all-in-one Tactically Responsive Space mission covering the rocket, satellite, and on-orbit operations. The firm also called the Iridium acquisition transformative and said it further validates Rocket Lab’s move toward becoming a vertically integrated, end-to-end space company. 10. Intel $INTC is investing $5.7B to expand manufacturing at its Leixlip campus in Ireland. The project will upgrade existing fabs, install new leading-edge equipment, and expand capacity for Xeon 6 and next-gen Xeon processors built on the Intel 3 node. Intel says the investment supports AI and high-performance computing demand while strengthening Europe’s semiconductor supply chain. The Leixlip site employs 4,900 people, with Intel’s total investment in Ireland now above €30B since 1989. 11. Fluence $FLNC signed a contract with Avantus to supply a 200MW / 800MWh Smartstack battery system for the Rexford 2 solar-plus-storage project in Tulare County, California. Construction is expected to begin in 2027, with operations targeted for late 2028, and enough capacity to power roughly 84,000 homes. 12. Bernstein says “this time is different” for memory $SNDK $MU, arguing new long-term agreements offer much stronger downside protection than past semiconductor contracts. Unlike prior cycles where customers delayed orders or disputed obligations, Bernstein says today’s memory LTAs are backed by upfront collateral. The firm estimates $SNDK has $11B+ of guarantees supporting roughly $69B of remaining obligations, while $MU holds $18B of deposits plus $4B of letters of credit. Bernstein also says protection is back-end weighted, with collateral coverage rising toward 75%–100% in the outer years when downcycle risk is highest. The bigger shift: counterparties are now hyperscalers and large OEMs with stronger balance sheets, while demand is tied to multi-year AI infrastructure buildouts and rising HBM/DRAM/NAND intensity. Bottom line: LTAs do not eliminate memory cyclicality, but Bernstein believes they improve earnings visibility, pricing durability, and trough protection versus prior cycles. WALL STREET IS THE GREATEST SHOW ON EARTH.

Meta is significantly scaling up its Louisiana Hyperion AI data center supercluster, transforming it into a massive 5 gigawatt powerhouse. • THE CAPEX BOOM: The updated project budget eclipses the original $27B joint venture estimate, with total AI capital expenditures now projected to cross over $50B • THE CASH ENGINE: Meta continues to expand its infrastructure directly from organic cash flows, backed by over $214.9B in total revenue and an elite 41.21% EBIT margin. • THE QUANT VIEW: While a 32.84% net margin showcases supreme financial health, the Seeking Alpha Quant system maintains a neutral HOLD as the tech giant absorbs massive near-term capital costs. Can $META out-compete big tech by funding its multi-billion dollar AI data infrastructure purely through organic cash flow?

Morgan Stanley: CapEx Major AI Infrastructure & Capex Surges > $1.4 Trillion Capex Forecast: Morgan Stanley has raised its total hyperscaler capex forecast to $1.4 trillion by 2028 (up from ~$1.2 trillion in 2027), driven by an updated bottom-up cost model and rising forward capacity expectations. > 4x Compute Capacity Growth: This heavy investment is projected to expand total available hyperscaler compute capacity to 120 GW by 2028, up from ~30 GW in 2025. > Rising Data Center Costs: Estimated data center deployment costs per gigawatt are increasing across multiple architectures due to inflationary pressures on memory and "outside the rack" (powered shell) development costs. Costs are modeled at $35bn for GB200, $39bn for GB300, and $49bn for Vera Rubin. > Bottlenecks and Timelines: Constraints on chips/racks, powered shells, and other supply chain bottlenecks are stretching data center construction timelines to as long as 3 years. Meta > Capex Expansion: Morgan Stanley significantly raised Meta's '27/'28 capex projections by 29% and 22% respectively, reaching $225bn in 2027 and $250bn in 2028. Higher depreciation from this spend modestly reduces '27/'28 EPS estimates. > API Revenue Stream: The rollout of Muse Spark 1.1 is priced aggressively (30%–85% below private peers). Morgan Stanley models that every 100 MW of compute allocated toward Meta's API could generate $8bn in revenue and ~$2 of '28 EPS. Amazon > Capex and Growth Expansion: Total company capex estimates were raised by 15% and 29%, reaching $308bn in 2027 and $318bn in 2028. > AWS Revenue Acceleration: Backed by private lab deals, AWS revenue growth is projected to hitting 40% y/y in '27 and 36% y/y in '28. AWS backlog is expected to hit ~$475bn in Q2. Google > Cloud Dominance vs. Tactical Risk: Google Cloud is modeled to achieve massive 77%/78% growth in 2026/2026. $GOOGL $NVDA $AMZN $META

Still can't believe investors were selling ai/dc stocks a couple weeks ago because they thought $META had excess compute.. obviously not true... since those FUD headlines $META has released new LLM models and increased their capex plans significantly (see below). btw, Morgan Stanley just increased their 2027 hyperscaler capex number to $1.2 trillion, up +58% from their 2026 capex estimate (they should have included $ORCL as well) I'll keep buying the dips in semis, memory, optics, connectivity and neoclouds... thanks for the pullbacks.

$META holding up very well today on a bad tape day.

Everyone’s worry is on AI capex slowdown. This shows it’s nowhere near done. $META just expanded its Louisiana data center to 5GW, taking the announced investment from $10B to over $50B. Same day, Morgan Stanley puts hyperscaler capex at $1.2 trillion in 2027 and $1.4 trillion in 2028. A month ago the market panicked that Meta had too much compute. Now they’re 5x-ing a single campus. Capex isn’t slowing down.

Meta is increasing AI data center spending, not lowering it. CNBC: Meta "said in a blog post on Monday that the site in Richland Parish, Louisiana — home to what will be Meta’s largest data center — will be a 5GW facility and cost over $50 billion. That’s higher than the $27 billion figure that was revealed in October"

$META just increased its announced investment in Richland Parish AI campus from $10B to over $50B, expanding the site to 5GW of compute capacity. Sounds like something a company with "excess compute" would do.

$META Meta Platforms Inc. has committed to spending an additional $40 billion on its data center campus in Louisiana, pushing its total expected investment beyond $250 billion for the site. The company will expand the project to at least 5 gigawatts of computing power at a cost of $50 billion, and has previously announced a $10 billion investment for the data center and surrounding community.

$META is expanding its Richland Parish, Louisiana data center to 5GW of compute capacity, lifting the announced investment to over $50B from the original $10B plan. Bloomberg reports the total site cost could top $250B when including chips, though Meta has only publicly disclosed $50B. Meta also plans $1B+ in local infrastructure improvements and says the site will support 1,000+ roles once operational.

$META looks like it put in a long term bottom. We could still see a quick rejection back toward 600, but this wedge is setting up for a push into new all time highs over the next 3 months. https://t.co/74XUxBTS2I

Just a semi recap TLDR: - $GLW glass bridge per Morgan Stanley has potential, but hard to displace FAU (like FOCI) in short term - $SPCX Starlink Gen 3 is scaling to 100,000 units (10x prev gen) creating possible capacity constraints for suppliers of switches to CCL. - PCB supply shortages are projected to persist until 2028, and component shortages/price hikes are already forcing ODMs like Inventec to issue conservative H2 shipment - DeepSeek and Zhipu are developing custom ASICs to bypass $NVDA (kinda expected by now). - Anthropic has achieved a $30B ARR and is projected to hit >$1B in Q3 profit. Turns out these frontier labs are more profitable than people think. - US admin pressured $AAPL to source from $INTC, in exchange for tariff exception. Kinda pressuring alignment away from TSM. - $TSM plans a 30x expansion of its Photonic Integrated Circuit (PIC) capacity by 2028, growing from 500 to 25,000 wafers per month - Hanmi Semiconductor is entering the CoWoS packaging equipment market - $NVDA and NTT hosting a conference July 24th to discuss CPO strategies. - 5x general NAND flash price hike this year, market size $489B by next year triggered by RAG/inference. Samsung and SK Hynix make emergency fab investments, NAND equipment suppliers go brrrr. - Gas turbine bottleneck, 40% supply gap accompanied by staggering 5-year delivery cycles (mega-fabs need them for mass production). - AI probe card assembly has hit a severe bottleneck apparently. Which is attempted to by solved with things like Innovation Service's machines. - Nanya, 79.5% gross margins reported, memory go brr. 4x capex for capacity/advanced packaging. - CXMT's STAR Market IPO on July 16, so should bring a lot of attention to memory players in that supply chain. - KYEC $1.4B investment into US for $TSM Arizona output test facilities. Lot of these players like $AMKR and others should go brr in 2028. - $INTC CEO warned last month that helium could hinder the manufacturing costs and delivery times of AI chips - 3D NAND word lines are shifting from Tungsten to Molybdenum starting at the 375-layer node - SambaNova secured JPM for AI inference and raised $1B at an $11B valuation. - HBM prices projections to double in 2027 as $NVDA Rubin platform drives demand. - $MU provides $500 million in funding to support GlobalWafers' US manufacturing capacity - $META 'Iris' will enter mass production in September via $AVGO and TSMC and Meta aims to double computing power to 14GW by 2027. - Largan Precision has secured its first CPO FAU order, mass production slated for middle of next year (kinda indication around Foci and others). - Samsung and SK Hynix have delayed the implementation of hybrid bonding packaging technology for HBM4 apparently - Memory costs (DRAM/NAND) have reached 60% of the BOM for sub-$400 smartphones, causing a severe volume contraction. - SK Hynix successfully rasied $26.5 billion through a Nasdaq ADR - $TSLA issued procurement guidelines requiring suppliers to reach weekly production of 1,000 units by September and double to 2,000-2,500 by year end for its 3rd gen Optimus robot. Apparently Alliance Technology and A-Link may be in this supply chain harmonic reducers and vision lenses? Kinda go through all this stuff every day, but don't wanna be a news reporter so just consolidated stuff I found interesting.

Wild "Amazon, Alphabet, Nvidia, Meta, Oracle and SpaceX, have issued a record $182 billion in investment-grade bonds so far in 2026. This marks a +1,300% increase from ~$13 billion over the same period in 2025. As a result, these 6 firms account for nearly 15% of total US corporate bond issuance year-to-date and over 50% of this year's growth in corporate bond issuance."

It’s crazy that $META is only a $1.7 trillion company when you list out what they actually own. WhatsApp. 3.3 billion users. The default messaging app for most of the planet. Bought for $19 billion, worth multiples of that now. Facebook. 3.07 billion monthly users. The largest social platform in human history, still growing. Instagram. 3 billion users. The default photo and video app for an entire generation, and one of the most valuable acquisitions ever made at $1 billion. Messenger. Another billion-user platform on its own. Threads. Youngest of the family and already past 400 million monthly users. Add it up and almost 4 billion people, half of humanity, touch a Meta app every month. Then the products stacked on top. The ad machine that prints over $160 billion a year and funds everything. Ray Ban Meta glasses, the first AI wearable that actually sells. Quest headsets. Meta AI pushed to billions of users overnight. Their own AI chip, Iris, entering production. And a compute buildout scaling to 14GW that rivals entire nations’ power grids. One company owns the attention of half the species and is converting it into an AI empire. $1.7 trillion might end up looking cheap.

AI is not a bubble. Lazy AI bears point to $NVDA's market cap and through PTSD, claim it resembles $CSCO in March 2000. However, any useful bearish analysis should look at what actually made the 1990s market a bubble from a macro sense, and whether those conditions exist today. I now refer you to the attached chart. In the late 1990s the two lines veered apart. Tech investment went vertical toward ~4.5% of GDP, while the economy-wide profit share rolled over from its 1997 high and fell hard into 2000. Investment surged while profitability eroded...bubble! Today, the lines rise in tandem. Tech investment has pushed to roughly 4.9% of GDP, above the dotcom peak and climbing more steeply, while pre-tax corporate profits sit near 14% of GDP. Meanwhile leverage has (mostly) stayed contained and the US current account deficit is shrinking. However, bears point to record levels of investment in isolation, choosing to ignore the growing profitability in addition. Are they dumb, or are they ignorant? Probably both. In the run up to March 2000, share prices rose and multiples exploded. The market paid more and more for each dollar of invisible earnings. This time, forward P/Es have barely moved even as share prices rocketed, because earnings expectations rose alongside them. For example, the Nasdaq 100 trades around 23x forward earnings, near its own 10Y average versus ~60x in March 2000. But...but....the market is so concentrated!!1!1! Yes, the ten largest S&P 500 companies account for ~40% of the index, above the dotcom peak. But those ten companies contribute around 30% of total market earnings, compared to under 20% in 2000, and trade at roughly a 50% premium to the rest of the market against a premium north of 100% at the prior peak. Now, bears will say: "If the rally is earnings driven, everything depends on whether the earnings persist!" Correct. But unfortunately for the bears, this is where things get uncomfortable. AI type names have added on the order of $27 trillion in market value since late 2022, up from roughly $19 trillion just seven months earlier. Set that against any weak attempt to discount the additional profit streams AI can plausibly generate for US companies (estimates cluster in the trillions) and the market has capitalised a multiple of the realistic prize. Not all of that $27 trillion is AI (the hyperscalers run enormous non-AI businesses), and more aggressive assumptions on adoption and productivity can lift the number. But closing the gap requires increasingly heroic assumptions: - that recent shifts in earnings shares are highly persistent - that the boom's suppliers capture an outsized slice of AI's total economic gains - that the economy-wide profit share keeps climbing indefinitely Alright, cool. But what about all the circular financing?! - Nvidia has committed tens of billions to OpenAI while remaining its primary chip supplier - OpenAI has signed a cloud commitment with $ORCL reported around $300B - Oracle in turn buys from Nvidia - $MSFT is simultaneously OpenAI's largest investor and one of its largest vendors True, this somewhat resembles the dotcom vendor financing where Cisco booked loans to cash stricken carriers as revenue (roughly a tenth of sales at the peak), much of it later written off. However, today's arrangements are mostly equity stakes in counterparties with genuinely fast growing revenue rather than disguised loans to fund purchases, and Nvidia has lately been unwinding parts of its ecosystem book. And according to analyst reports, even the AI labs like Anthropic have now turned profitable. A feat many thought would be impossible only a year ago. Personally, I treat this circularity as risk rather than a point to build a bear case around since it's all ultimately leading to greater earnings across the board. Even for fronteir labs. Shock! This then leaves the one key question: Will barriers to entry protect today's profits from erosion? This entirely depends on each company's position in the AI supply chain. - At the model layer, barriers are relatively fragile where frontier models will almost certainly converge longer-term, and open source alternatives have the ability to reset price floors. However, AI soverignty will ultimately result in the likes of OpenAI/Anthropic winning. - At the hyperscaler layer, $AMZN, $GOOGL, $META, and $MSFT are set to spend (currently) $750B for 2026 AI capex where falling behind is not an option. These companies are led by people smarter than you or I - do you think they'll risk their entire business collapsing for AI? No. In fact, you can already see that AI is boosting their earnings measurably in recent earnings. - Going further down, you've got irreplaceable companies such as $ASML (EUV machines), $TSM (CoWoS packaging), and HBM with $MU, SK Hynix and Samsung who are gated by long qualification cycles and multi year LTAs where demand > supply up to the 2030's. The risk of a 2000 style valuation bubble is massively lower than the bearish consensus believes. The world is revolving around AI, and that'll continue for the forseeable future.

Meta CTO shared that Meta's objective remains capturing long-tail data sets of human mouse movements to teach AI how to natively navigate legacy digital silos. Despite pausing their controversial employee-tracking program due to an internal data leak. $META https://t.co/eoxlvwtt44