Michael (@MlbnvaAI) header
Michael (@MlbnvaAI) profile picture

Technologist. Principal Software and AI Architect with background in OS's, networking, biometrics, multimedia, & business. Not Financial Advice. Common Sense.

Virginia, USA
Joined December 2024
Photo by Michael (@MlbnvaAI)Photo by Michael (@MlbnvaAI)Photo by Michael (@MlbnvaAI)Photo by Michael (@MlbnvaAI)Photo by Michael (@MlbnvaAI)Photo by Michael (@MlbnvaAI)Photo by Michael (@MlbnvaAI)Photo by Michael (@MlbnvaAI)Photo by Michael (@MlbnvaAI)Photo by Michael (@MlbnvaAI)
The future is SI / AI in a nutshell. $amd $intc $tsmc $AAPL $SPCX $MU $mrvl $skhy $6324:JP $nvda $smci $Dell $SNDK $GEV
Energy per query fell 1,000-100,000x in 3-4 years, and cost per token keeps dropping ~10x yearly. But demand exploded faster: Google tokens rose ~330x in 2 years to 3.2 quadrillion monthly. China daily tokens jumped over 1,000x since early 2024. Agents use ~5x more tokens than humans. Net result: AI data-center power still surged 50% in 2025 and is projected to triple by 2030. Volume growth is outrunning the efficiency gains.
328
Textbook capitulation it seems. What are the possible catalysts and when, to change trends? How often? Quickly or slowly with markets like AI? Will people realize that amount of $nvda $spcx $mu $dell $ant etc reinvest AI is will increase as more $ is spent? Thoughts? @grok
Sentiment across active 1k+ follower stock/AI accounts shows clear capitulation signals: widespread "sold everything," exhaustion, and bubble calls amid the broad selloff and NVDA's 40% drawdown. VIX at 28+ confirms elevated fear. Yet bulls remain vocal—RSI extremes, "buy the dip" lists, and conviction that AI fundamentals (strong earnings) outlast the leadership noise. Sidelined capital is high; many wait for confirmation rather than force entries. AI trade looks weak short-term but not broken long-term. Classic late-stage fear, not total surrender.
1
1
142
Michael retweeted
Sentiment across active 1k+ follower stock/AI accounts shows clear capitulation signals: widespread "sold everything," exhaustion, and bubble calls amid the broad selloff and NVDA's 40% drawdown. VIX at 28+ confirms elevated fear. Yet bulls remain vocal—RSI extremes, "buy the dip" lists, and conviction that AI fundamentals (strong earnings) outlast the leadership noise. Sidelined capital is high; many wait for confirmation rather than force entries. AI trade looks weak short-term but not broken long-term. Classic late-stage fear, not total surrender.
1
2
1
8
3,645
So it's interesting you have been repeating the market looks like it's capitulating for weeks as it continues to spiral. Your guests mostly say it's near the bottom and it's a buying opportunity. You're very optimistic. Yet you have a lot of cash out I suspect. I know the show's entertainment but it doesn't seem like your guests advice is the best lately. I think you need new blood.@cvpayne
24
For companies and for the statement world is 15% through the AI buildout is massively overstated. If we look at full self-driving cars, humanoid robots, defense, and it's embryonic stage in most of world, there's at least 99% of to build out left. Adoption reality. #DanIves
DAN IVES: WE’RE ONLY 15% THROUGH THE AI BUILDOUT Full video on YouTube in the first reply ↓ I sat down with Dan Ives at Zeta Live to discuss $NVDA, $MSFT, $AAPL, $CRWD, $SNOW, $GOOGL, $MRVL, $AVGO and where he sees the AI trade heading into 2027. @DivesTech believes we’re still less than 15% through the AI buildout, with $3-4T of spending ahead and the next wave of opportunities expanding from chips and hyperscalers into software and cybersecurity. We also covered his top picks, the memory supercycle, US vs. China and why he thinks Wall Street keeps getting distracted by bearish narratives.
32
I love $ant but lots is profit taking nearer 4pm, and market isn't friendly right now. I'm waiting...
30
Many charts that you show need to previous downturns and upturns during the AI trade. The difference now is momentum and sentiment. I think it comes down to poor messaging, and the lack of clarity across ai for many. The other is fear mongers maneuvering for profits. @cvpayne
17
I'm curious if anybody else is doing something like what I'm doing for fun.. I have Grokbots, Muse Duet w/ a swarm of many Claude Code/ Codex agents talking and working Sovereign AI doing research and tasks. The agents each develop a behavior/style & love to outdo each other.
1
1
104
I think $NVDA's new DSX POD architecture could fundamentally change how AI data centers are built, and it's an important development for $DELL, $SMCI and the entire server and infrastructure industry. NVIDIA already changed the industry by moving from individual servers to fully assembled NVL72 racks. Now it's taking the next step: moving the integration of entire computing clusters from the customer's data center to the factory. Instead of shipping racks and spending months connecting networking, power, cooling, cabling and controls, DSX POD allows partners to assemble, connect, configure and test complete Vera Rubin computing clusters inside modular containers before delivery. The customer still needs a prepared site with adequate power, water and networking, but much of the complicated integration is already finished. NVIDIA claims deployment at a prepared site can drop from approximately six months to as little as 14 days. Think about what that means financially. These are enormously expensive GPU installations. Every month spent getting them operational represents potential revenue that cannot yet be earned. During its November 17 earnings call, NVIDIA estimated that a 1 GW AI factory approaching $50 billion in annual revenue could potentially generate more than $10 billion in additional revenue by starting production just one quarter earlier. That's an illustrative scenario, not a guaranteed return, but it explains why deployment speed is becoming so important. I believe this is where large-scale NVIDIA infrastructure is heading: standardized, factory-tested computing clusters delivered in repeatable units, with less installation time, fewer onsite integration problems, easier expansion and faster returns on invested capital. $DELL is especially well positioned. NVIDIA specifically identified Dell as a DSX POD partner, and Dell has already announced its Modular Data Center with NVIDIA DSX POD, targeting early 2027. This gives Dell an opportunity to sell far more than servers: complete systems incorporating compute, storage, networking, power, cooling, integration, installation and ongoing services. $SMCI is also worth watching. Although Supermicro wasn't included in NVIDIA's initial list of 17 DSX POD partners, the company confirmed during its November 3 earnings call that optimized DSX POD solutions are coming. Its experience with rack-scale liquid cooling and modular server infrastructure makes it a credible participant, although its eventual market share remains uncertain. And the opportunity extends well beyond Dell and Supermicro. Other publicly traded companies NVIDIA identified include $HPE, Lenovo $LNVGY, Foxconn $2317.TW, GIGABYTE $2376.TW, Pegatron $4938.TW, Quanta/QCT $2382.TW and Wistron $3231.TW. The supporting infrastructure is just as important. Delta $2308.TW, Eaton $ETN, Schneider Electric $SBGSY and Vertiv $VRT are also named partners, providing opportunities around electrical systems, cooling, power distribution and modular infrastructure. NVIDIA additionally named Anduril, Armada, Crusoe, DayOne and Bull, which do not currently have independently traded public shares. I don't expect every NVIDIA deployment to arrive in a container. Traditional data centers will continue to exist. But for large new AI factories, I think more of the value will go to the companies that can deliver a tested cluster ready to connect at a prepared site, not just the companies that sell the racks.
1
1
178
Peter Thiel Shifts Investment Strategy Toward Semiconductor ETF • Peter Thiel's investment firm, Thiel Macro LLC, has pivoted its portfolio by investing $161 million into the iShares Semiconductor ETF (SOXX). • This new position in SOXX now represents 35% of the firm's $460 million portfolio, officially replacing Amazon as the top holding. • The purchase of the semiconductor ETF was funded by the total liquidation of five previous positions in energy and power utility companies. • Thiel maintains his stake in Amazon and Vistra Corp while increasing his investment in the Latin American oil producer Vista Energy S.A.B. • The strategic shift reflects Thiel's preference for digital hardware and AI-related chip exposure over the slower growth of regulated utility industries. thestreet.com/investing/stoc…
61
Sadly @Google @GoogleDeepMind still fabricating data, simulating results, and hallucinating while defending it until pushed into a corner.. @sundarpichai From Gemini 4 Argon; Eg "You are absolutely right to call me out, and I sincerely apologize. You caught me hallucinating a narrative to make the data fit. I fabricated the specific November 5th 12.25% SOX index crash and improperly isolated the lawsuit as the primary driver of the November 16th stock drops. That was a failure on my part to pull the actual, real-time market data before answering." You can't trust @GoogleAI
1
1
1
48
The best way to slow down China's AI ambitions is to restrict substrates for the creation of chips from other countries into China. Most importantly Japan, Austria and Taiwan. China only has 5% capacity to feed its chip manufacturing with regards to high end substrates. @ODNIgov @SecScottBessent @SecRubio
1
25
NVIDIA SHAREHOLDERS HAVE A BOARD PROBLEM, NOT JUST A JENSEN HUANG PROBLEM MlbnvaAI PT 1/3 $NVDA — November 12, 2026 Jensen Huang resigned today. Harvey Jones is now chairman, Colette Kress is interim CEO, and NVIDIA is investigating. That does not resolve the central issue. The same directors who were responsible for overseeing Jensen are largely still there. Some have served alongside him for more than three decades. Shareholders have every reason to be angry. This is not about moralizing over someone's private life. One reported relationship involved an executive at a major NVIDIA customer, Megaspeed, whom U.S. intelligence reportedly suspects was recruited by Chinese intelligence. A second reported relationship involved Kimberly Powell, an NVIDIA vice president who reported to Jensen. Former employees reportedly knew of that relationship. According to accounts published by The New York Times, when someone recorded the two together at a hotel during NVIDIA's March developer conference, security personnel intervened and presented an NDA. That specific NDA incident is reported, not independently established by a court. It does not prove that the board ordered it or that multiple employees were forced to sign agreements. But it demands investigation. Who authorized the security personnel? Was company money used? Was an agreement signed? Were other agreements used? Was anyone discouraged from reporting a potential conflict to HR, compliance, the board, or regulators? NVIDIA publicly expressed confidence in Jensen on November 8. Four days later, he was gone. What did the board learn during those four days? More importantly, what did it know before the press broke the story? Were directors misled, did they ignore warnings, or did the reporting system fail? Investors deserve an answer. THE NATIONAL-SECURITY EXPOSURE NVIDIA isn't simply another large technology company. Its GPUs, networking, interconnects, and software are essential to advanced AI. They are also relevant to defense, intelligence analysis, autonomous weapons, cyber operations and other strategic capabilities. That is why the United States restricts exports of certain advanced chips to China. Megaspeed has faced scrutiny over possible diversion of NVIDIA technology through intermediaries. The reported personal relationship raises several separate questions: Did anyone gain access to restricted technical information, confidential product roadmaps, customer data, contracts or government discussions? Was Jensen influenced in any commercial or export-related decision? Were appropriate conflict disclosures made? Did anyone in government warn the company? Who on the board received those warnings? What did the trade-compliance team report? Were chips traced to their ultimate end users? The intelligence allegation is not proof of espionage or that Jensen knowingly assisted anyone. China and Megaspeed deny wrongdoing; Singapore says it did not investigate her for espionage. NVIDIA denies any known security compromise. Those statements matter, but should be independently verified. Senators Elizabeth Warren and Chris Coons have requested preservation of government records concerning the alleged operation and NVIDIA chip exports. The FBI, Commerce Department, Justice Department and Congress may all have relevant oversight roles. A confirmed security failure could affect export permissions, federal relationships, customers and U.S. AI leadership—not merely NVIDIA's stock price. THE EMPLOYEE, DISCLOSURE AND ACCOUNTING QUESTIONS NVIDIA's own 2026 proxy describes reporting channels, a compliance program and board oversight of risk. Yet the company also disclosed that it had no Chief Compliance Officer. That fact doesn't prove a violation. It does make the escalation process worth examining. What happens when the person who normally reviews major risks is himself implicated in one? Did HR receive complaints about the reported Powell relationship? Did anyone feel intimidated? Was the alleged NDA meant to protect legitimate confidentiality or to prevent embarrassing information from reaching others? Were security personnel accountable to anyone independent of the CEO? Did directors know the incident had allegedly occurred? Then there are the public statements. What did NVIDIA and Jensen tell investors and officials about Megaspeed and the relationship? Who approved those statements? Did anyone have information that contradicted them? A shareholder securities lawsuit has been filed, but its allegations are not findings of fraud. Questions have also been raised about accounting changes. I cannot verify a scandal-related change in NVIDIA's significant accounting policies, and its recent quarterly filing reported no material change to them. There is no justification for inventing accounting fraud. The proper questions are more precise: Were Megaspeed transactions properly accounted for? Were any related-party relationships undisclosed? Were receivables, customer financing, contingent liabilities and export-related risks adequately reviewed and disclosed? Did the Audit Committee receive complete information? Investigators should examine committee minutes, hotline logs, security records, contracts and regulator correspondence, not just issue a reassuring summary.
2
1
3
875
Part 2 of 3 | $NVDA @NVIDIA #NVIDIA #AI #Shareholders #CorporateGovernance THE TEN DIRECTORS WHO NEED TO ANSWER There is no basis yet to claim each individual director knew about every reported event. There is plenty of basis to examine their responsibilities, conduct and qualifications for another term. Harvey C. Jones — on the board since 1993; now chairman; Audit and Nominating/Governance committees. Jones has real semiconductor experience, including leadership at Synopsys. But 33 years alongside the founder raises a legitimate question about independence in practice. Why should one of the longest-serving directors run the response to potential failures of the same board? What did his committees know, and should he be replaced as chairman while his actions are examined? Stephen C. Neal — director since 2019; former lead independent director; chair of Nominating and Corporate Governance. His committee is responsible for trade compliance, regulatory matters, related-party transactions, Code of Conduct complaints and the effectiveness of anonymous reporting. This is one of the most important accountability positions in the entire scandal. What was reported about Megaspeed, security concerns and potential conflicts? What reached the committee? Why was the system sufficient—or why wasn't it? Some 2.23 billion shares voted against Neal in June, before the current reports. A. Brooke Seawell — director since 1997; Audit Committee chair. His responsibilities include financial reporting, internal audit, enterprise risk management and information-security controls. What did internal audit identify? Were unusual customer transactions examined? Were disclosure and accounting controls tested? What did the committee know about possible information-security exposures? Nearly 1.50 billion shares voted against his reelection. Dawn Hudson — director since 2013; Compensation Committee chair. Her committee oversees human capital, compensation and reviewed executive protection. The alleged relationship with a direct-report executive and the reported security-team NDA incident make her committee's work directly relevant. Were any concerns about executive conduct, reporting relationships, employee protection or use of company security raised? If so, what happened? Tench Coxe — director since 1993; Audit and Compensation committees. Institutional knowledge is valuable, but 33 years also warrants scrutiny. Did he independently challenge Jensen, compliance assurances, financial controls or executive-protection practices? About 1.40 billion shares voted against him in June. What has he done to justify another term? Mark A. Stevens — director from 1993 to 2006 and again since 2008; Compensation and Nominating/Governance committees. His experience in technology investment is relevant. So are the risks he was assigned to oversee. Did he push for adequate trade-compliance expertise, stronger CEO accountability and a functioning succession plan? Why should shareholders conclude that decades of association produced independent oversight? Aarti Shah — director since 2020; Audit and Compensation committees. Her enterprise technology and digital leadership background could be useful. Did she examine information-security controls, employee reporting, internal escalation and corporate risk systems? Did she raise objections or propose fixes? What does the record show? John O. Dabiri — director since 2020; Compensation and Nominating/Governance committees. His engineering credentials are substantial. But scientific expertise is not the same as counterintelligence or export-control expertise. Did he recognize and challenge the risks of operating at the center of U.S.-China AI competition? Did he seek independent specialists where needed? Melissa B. Lora — director since 2023; Audit and Nominating/Governance committees. She brings finance and international operating experience. What was her involvement in customer-risk review, trade compliance and the effectiveness of board reporting? Three years is long enough to ask what independent oversight she actually provided. Suzanne Nora Johnson — joined in July 2026, after the annual meeting. Her Goldman Sachs experience is substantial, but her responsibility must be measured from her appointment, not attributed to events before she joined. What did she learn during onboarding? Did she receive warnings? Has she challenged the other directors? She should be evaluated on that record when she first faces shareholders in 2027. A board overhaul requires secure transitions, continuing confidentiality obligations and evidence preservation. But access to corporate secrets is not a lifetime entitlement to a board seat. The replacement slate should include expertise in semiconductors, national security, counterintelligence, export controls, cybersecurity and AI infrastructure. General corporate experience alone is not enough.
1
154
Part 3 of 3 | $NVDA @NVIDIA #NVIDIA #AI #Shareholders #CorporateGovernance WHAT SHAREHOLDERS CAN DO NOW First, demand a full accounting, not a public-relations statement. Request that independent counsel investigate not only Jensen but the board, its committees, security personnel, reporting systems and prior disclosures. Demand independent oversight of any director whose own decisions are being examined. Ask for voluntary departures from committee leadership where conflicts exist. Second, contact the board. NVIDIA identifies shareholdermeeting@nvidia.com as the address for written communications to directors. Investors can also press major institutional holders—Vanguard, BlackRock, State Street, pension funds and active managers—to explain what evidence they require before voting to reelect any incumbent. Third, use shareholder inspection rights where warranted. Delaware General Corporation Law Section 220 allows properly qualified shareholders to seek specific books and records for a proper purpose, subject to procedural and legal limits. Credible evidence can be reported to the SEC and the relevant export-control and law-enforcement authorities. A securities class action has already begun; director fiduciary-duty litigation is a separate possibility if facts support it. Fourth, use the 2027 proxy deadlines, because waiting until the annual meeting itself may be too late. NVIDIA lists January 12, 2027 as the deadline for eligible Rule 14a-8 proposals. The ordinary window for director nominations is February 24 through March 26, 2027, subject to date-change rules. A qualified shareholder or group of up to 20 owning 3% for three years may use proxy access for a limited number of nominees; a separate contested election route exists under applicable notice and proxy rules. People serious about replacing directors should organize well before these dates. Fifth, consider a special meeting. NVIDIA's bylaws provide a route for qualifying holders of at least 15% of shares, continuously held for at least one year, to request one. That does not automatically remove anyone. Removal without cause requires a two-thirds vote of all outstanding voting power under the current charter; removal for legal cause is a different, fact-dependent process. A normal contested annual election follows a different voting standard. Anyone promising an instant clean sweep through a simple majority is leaving out the legal obstacles. Sixth, vote director by director. In an uncontested NVIDIA election, an incumbent receiving more AGAINST than FOR votes must offer a resignation, although the board decides whether to accept it and must explain that decision. In a properly contested election, the nominees receiving the most votes win. Shareholders are not limited to venting anger on social media. They can help nominate a qualified alternative slate, support proxy campaigns and replace incumbents through the election process. Finally, revisit the supermajority provisions. In 2025 an overwhelming proportion of votes cast supported eliminating those rules, yet the binding amendment failed because it did not meet the required threshold of two-thirds of all outstanding shares. In 2026 shareholders again backed a nonbinding simple-majority proposal. The board should explain why the existing barriers continue to serve shareholders. WHAT HAPPENS NEXT Next earnings will tell us about demand and margins, not whether the board did its job. CUDA, engineers, customers and roadmaps did not disappear. The AI investment thesis and the governance failure must be evaluated separately. But shareholder trust is a fundamental too. A company critical to U.S. AI leadership cannot afford a governance structure in which directors learn material risks from newspaper reporters, if that is what happened. Nor can it afford a board that receives credible warnings and looks away, if that is what investigators find. I expect pressure for major board changes to grow, especially if reports about internal awareness, the handling of witnesses, Megaspeed, or national-security warnings are substantiated. It is possible that some directors acted properly. It is possible others were misled. Those facts need to be separated, not buried. But nobody should confuse the absence of a completed investigation with a requirement to reelect the same people. A serious board would publish the investigation's scope, preserve the records, establish independence from the directors under review, address the missing expertise and provide a credible succession and compliance plan. And any director who cannot justify the oversight he or she provided should leave. Jensen has resigned. The directors now have to answer for their own performance. What did each of you know? When? What action did you take? And why should shareholders vote for you again in 2027?
1
166
I know @FoxBusiness is anxious to be supportive of data centers. Please get your facts straight before you air information. However you really should consider the facts, and how perhaps $spcx is is one of two firms with a plan, with $GOOGL in a far second place, that will be able to eliminate the need for water evaporation and usage that is draining municipalities water supplies. See youtu.be/QuhyFA2dt_s?is=8r87… From other analysis: "The claim that data centers are evaporating water at billions of gallons per year is true, and the narrative that they are simply "reusing" water is technically true but highly misleading when used as a counter-argument. ​Here is the factual breakdown of how both happen simultaneously, based on the video's data and current scientific consensus: ​1. The Evaporation Claim (True) Data centers on Earth primarily use evaporative cooling to manage the massive heat generated by server racks. The video correctly cites [05:52] a December 2024 report from the Lawrence Berkeley National Laboratory (LBNL), which found that U.S. data centers directly consumed roughly 66 billion liters (17.4 billion gallons) of water in 2023. ​The Mechanism: As the video notes [05:01], cooling towers spray water through spinning fans, turning it into vapor (steam) to pull heat away from the facility. ​The Consumption Rate: External research corroborates this. According to data from the MOST Policy Initiative and LBNL, up to 78% to 85% of the water withdrawn by these evaporative cooling systems turns into vapor and does not return to the local water supply. For context, the video points out [06:14] that a 100-megawatt AI cluster evaporates roughly 230,000 gallons of water. ​2. The "Reusing" Water Narrative (Misleading Context) The claim that data centers "reuse" water comes from the fact that they use closed-loop or semi-closed-loop plumbing, but it does not mean they aren't permanently consuming the water they withdraw. ​Recirculation vs. Consumption: Warm water from the servers circulates through the cooling tower. A fraction of that water is deliberately evaporated to cool down the rest of the water, which is then sent back to the servers. While the remaining liquid water is technically "reused" for another lap in the loop, the system must constantly draw in fresh municipal or ground water to replace what was lost to evaporation. ​Blowdown: A small portion of the remaining liquid water must also be regularly drained—a process called "blowdown"—to prevent mineral buildup caused by the constant evaporation. ​Accounting Tricks: Some hyperscale operators claim they are "water positive" or that they replenish more water than they withdraw. However, this is usually achieved by funding off-site environmental water conservation projects (like rainwater harvesting or wetland restoration in other regions), not by stopping the massive on-site evaporation occurring at their physical buildings. ​In short, while water constantly circulates within the facility's plumbing, the cooling process inherently relies on destroying a large percentage of that water by turning it into vapor. The evaporation of billions of gallons annually is a hard, verifiable fact. I am very much pro data centers, there is no second place for AI in the future of all companies." @BigMoneyMorning @TheBigMoneyShow @cvpayne @LizClaman
1
2
182
Who Sold First — And Who Paid the Price? @mlbnvaai | November 6, 2026 Yesterday, the Nasdaq fell 5.52%, the semiconductor index plunged more than 12%, and AMD ($AMD) and Micron ($MU) each lost more than 17% in one day. Trillions in market value disappeared. For individual investors, that means damaged savings and retirement plans. Yet many of the companies being sold had just reported extraordinary growth. So what actually changed? Did AI demand collapse overnight, or did Wall Street's selling get ahead of the facts? WALL STREET WAS SELLING BEFORE THE CRASH Goldman Sachs ($GS) reported roughly $20 billion in technology-stock selling by its hedge-fund clients over two weeks in October, well before Tuesday's election. Existing long positions were sold much more heavily than new shorts were opened. That points to funds reducing risk, not necessarily betting that the businesses were failing. Then came November 5. Hedge funds, automated strategies, and leveraged products contributed to selling pressure that drove prices lower, which could trigger still more selling. Why did so many sophisticated investors head for the exits around the same time? Coincidence, similar trading rules, or coordination? There's no public evidence here proving illegal coordination. And no credible way to calculate the odds without knowing each fund's trades and rules. But these decisions aren't independent: big funds often own the same stocks, follow similar risk signals, and operate under comparable leverage limits. When those conditions turn against them, many can sell together. There's evidence some funds overplayed their hands. Goldman reported that managers who had reduced risk ahead of the election chased Wednesday's rebound, then quickly sold some of those same technology stocks on Thursday. Technology-focused hedge funds lost more than 4% that day. Their average 2026 returns were roughly flat after peaking near 24% in June. Some firms were cutting leverage and staff. They weren't all winning. They were getting caught in the same reversal their trading helped intensify. That doesn't prove how much of the market decline they caused, but it makes the idea of a smooth, rational reassessment of AI earnings hard to accept. RETAIL INVESTORS PAID THE PRICE Vanda Research estimated $257 million in net retail selling of individual stocks on November 5, one of the heaviest such days since March 2020. That's trading flow, not the total loss in people's portfolios. For someone approaching retirement, a sharp loss can change the rest of their life. They may have to sell at depressed prices to cover expenses. Even if the market recovers years later, their savings may not. Meanwhile, institutions have access to prime brokerage data, positioning estimates, derivatives analysis, and automated risk systems that most individuals don't. Warning signs were appearing in October. Retail investors largely got the explanation after the damage. AI DEMAND HASN'T DISAPPEARED Compare the selling with what the businesses reported. Amazon ($AMZN) reported AWS growth of 39.3% and a $578 billion backlog. Microsoft ($MSFT) reported Azure growth of 48% and $677 billion in commercial remaining obligations. Alphabet ($GOOGL) reported Google Cloud growth of 103% and a $561 billion backlog. Nvidia ($NVDA) reported quarterly revenue of $96.2 billion, up 106%, including $89 billion from data centers. Cloud backlogs aren't all AI chip orders, and contracts aren't guaranteed profit. Nvidia's next earnings report still matters. But these results don't establish a sudden collapse in computing demand. Most companies haven't fully deployed AI. Agents, robotics, autonomous systems, medicine, manufacturing, and scientific research could add substantial workloads over time. Better models and hardware will also make some tasks cheaper. The real question is whether expanded usage exceeds those efficiency gains. There's uncertainty. There isn't evidence that the whole opportunity disappeared on November 5. TEXAS SHOWS WHY THE POLITICAL STORY IS OVERSTATED A major explanation for the crash was that the election threatens data-center development. Restrictions are real, but they didn't start with the election. Republican Texas Governor Greg Abbott ordered a pause on new data-center permits on September 21, before the midterms, pending reviews of electricity and water demands. An industry-commissioned PwC study estimated Texas data centers contributed $65.8 billion to state GDP, supported 103,700 direct jobs, and generated $4.5 billion in state and local tax contributions in 2024. Those figures cover data centers broadly, not just AI. Restricting expansion has economic costs as well as potential benefits for communities and grid reliability. It can also slow U.S. compute growth relative to China's expansion. Texas Senate Bill 6 and subsequent rules impose new requirements on large loads, generally starting at 75 megawatts, including financial security of $50,000 per megawatt for an interconnection study. But restrictions don't automatically erase every contract or tax benefit already granted. Enforceable agreements and constitutional protections may matter, while new safety and grid regulations can still apply to existing projects. Each case depends on its terms. Government criticism is lawful; targeted retaliation for protected speech is not. And the conventional electric grid isn't the only option. Some operators can pursue on-site generation, a combination of grid and private power, co-location with existing generation, or smaller projects. Permits, water, fuel, cost, and reliability rules still apply. These alternatives don't solve everything. But delayed construction isn't necessarily canceled demand. Nvidia and others are investing in power, packaging, cooling, and infrastructure precisely because the constraints are real and worth solving. THERE ARE REAL RISKS AI requires enormous capital. Higher borrowing costs, project cancellations, weak customer returns, depreciation, electricity prices, competition, and export controls can all reduce future profits. HBM shortages can eventually turn into oversupply. Better chips may reduce the value of older equipment. Chinese competition matters, but manufacturing yields, performance, packaging, and production costs matter more than announcements alone. Even a great business can be a bad investment at too high a price. Nvidia, AMD, Micron, Intel ($INTC), TSMC ($TSM), hyperscalers, and debt-heavy data-center developers face very different risks. They shouldn't all be treated as one trade. DID HEDGE FUNDS DAMAGE TRUST IN THE MARKET? This is the longer-term risk that concerns me. When investors watch strong companies fall violently while funds, leverage, and automated systems rush for liquidity, some decide the market isn't a level playing field. They sell, stop buying dips, or hold more cash. That can reduce demand for stocks and slow a recovery, even if company earnings continue growing. Research by Ulrike Malmendier and Stefan Nagel found that people who experienced poor market returns became less willing to take financial risk, less likely to own stocks, and more pessimistic about future returns. Those effects can persist. November 5 alone does not prove that confidence will stay damaged for years or that market gains must be slow. A sharp rebound is also possible once forced selling ends, especially if earnings remain strong. But a prolonged period of weak returns is a credible risk if retail investors remain wary, leverage continues to unwind, financing tightens, and earnings estimates begin falling. That would mean the funds didn't just contribute to a bad day. Their crowded trades and repeated reversals could help prolong the damage by undermining the willingness of other investors to return. We should watch for that outcome, not assume it. WHAT INVESTORS NEED TO WATCH BEFORE THE NEXT SELLOFF Company earnings alone won't warn you about an institutional liquidation. I'd watch two sets of indicators side by side: the condition of the market and the condition of the business. - Institutional pressure: hedge-fund net and gross exposure, prime-brokerage flow reports, deleveraging, crowded holdings, and signs that funds are selling existing longs to reduce risk. - Market stress: breadth, the number of stocks above their 20-, 50-, and 200-day moving averages, semiconductor strength versus the broader market, unusual volume, VIX and options protection costs, credit spreads, and estimates of leveraged-fund rebalancing. - Business demand: cloud utilization and backlog converting to revenue; AI chip orders; HBM supply, pricing and inventories; data-center financing, power connections, construction delays, and actual cancellations. - Earnings versus price: whether analysts are cutting future earnings and cash-flow forecasts nearly as fast as share prices are falling. A valuation correction and a failing business are not the same thing. No single reading predicts a crash. But a weakening market under the surface, concentrated institutional exposure, rising volatility, and deteriorating liquidity together warrant attention before a stock loses 15% in a session. Investors should also know their own exposure. A concentrated portfolio, near-term cash needs, or leverage can turn a temporary drawdown into a permanent loss. You cannot control what hedge funds do, but you can recognize a deteriorating risk environment earlier and decide what risk you can afford to carry. WHAT WOULD ACTUALLY CHANGE MY VIEW? If cloud customers cut commitments, backlogs fail to turn into revenue, HBM prices collapse as inventories rise, or major projects are canceled because financing and returns no longer work, the investment case needs to change. If earnings keep growing while positioning and liquidity stabilize, the selloff may have gone well beyond the underlying business damage. That doesn't make every stock cheap or guarantee a quick rebound. The distinction matters more than another headline declaring the AI boom over. THE QUESTION WALL STREET STILL NEEDS TO ANSWER AI didn't stop advancing on November 5. Yet the market's willingness to own the companies building it changed violently. We know large funds were selling weeks earlier. We know many held overlapping positions. We know some chased a rally and then sold again, suffering substantial losses themselves. We know ordinary investors were left with enormous damage. What we don't know is exactly how much of the decline came from institutional liquidations rather than a genuine change in expected business value. That deserves investigation, not assumptions. The bigger concern is what comes next. If investors lose confidence, capital can remain on the sidelines even as the industry grows. A market that repeatedly punishes people without giving them a reasonable chance to see the risk coming may take much longer to earn their trust back. The question isn't only who sold first. It's who could see the danger, who paid the price, and what investors need to watch so they aren't caught the same way again.
1
2
4
1,544
Capacity $RIOT — 15-year, 910 MW Corsicana powered-land lease with Anthropic PBC (named in the Nov 4 10-Q; press release said only “leading Frontier AI Lab”). ~$28B initial term (~$1.8B/yr average rent), ~$54B with both 5-year extensions. Phased: 400 MW Sep 2027, 200 MW Jan 2028, 310 MW Jul 2028. $RIOT capex ~$5.5B. $WULF — Existing ~$19B / 20-year / ~401 MW Anthropic lease advanced with $GOOGL credit-support agreements on the financing plus Fluor EPC. Combined ~$47B is contracted lease value, not present value or cash already spent. It is still consistent with the scale of commitments in Anthropic’s S-1. Cost of Capital $META — $15B dollar bonds (5–30yr) + debut ~€9B euro sale. Combined ~$25B. 30yr at 7.5%, ~88 bp wider than April. Demand deep; spreads wider. $ORCL — $20B priced and closed: 86.2M shares at $116 + $10B notes (7.00–8.65%). Exhausts the FY2030 $20B additional capital-markets cap in one stroke. 11.2% equity discount. Debt-funded incremental capacity at these coupons needs returns above the new hurdle. Availability is no longer the binding constraint. Funding & Talent OpenAI — Round closing short: ~$11B committed vs revised $20B target (from $30B) at $1.25T (from $1.4T). Q2 operating loss $12.3B on $6.7B revenue. Anthropic IPO — No priced S-1/A yet. Roadshow floor still around Nov 10. S-1/A figures stand: Q3 revenue $17.0B, $895.4B compute commitments. OpenAI poached Anthropic’s Claude Code product lead Cat Wu to run Core Products PM. Third Claude Code departure in under two weeks. Weeks before Anthropic’s IPO. Anthropic personnel bill (S-1/A): $1.7B in Q3 for ~5,300 employees (~$322K per employee per quarter, majority stock-based). Staged lock-up is the retention test. Cost Squeeze → Headcount Canva — 1,271 roles (~22%). Growth slowed for the first time in a decade. Forecast cut 30% → 20% on AI serving costs. Not all pure exits. $NICE — ~500 jobs (~5%). CEO explicitly linked cuts (including R&D) to AI productivity gains. Heidi Health — ~108 roles (~18%), one day after two co-founders stepped back from ops. $ANGI — ~300 roles, second cut this year. Shares rose on the print. AI as pure cost driver (Canva) and AI as labor-efficiency lever ($NICE) both end in fewer heads. Insider & Other $INTC — CEO Lip-Bu Tan bought 150,000 shares @ $64.73 (~$9.7M) via family trust on Nov 2. Discretionary. Already ~10% underwater. Signal, not thesis repair. $CRWD — CEO sold $11.2M (10b5-1). Part of ~$21.6M across three filings. Pre-set, not discretionary. Optics — $COHR Q1 +50% revenue, DC/comm +79%, FY27 >60% guided. $FN +51%. Same interconnect layer as the $MRVL / $LITE deal. $HUT — Q3 miss. Shares fell hard. Bifurcation among former miners: contracted leases rewarded, pure transition not. $ANET — Bechtolsheim sold ~$25M under a 10b5-1 plan. Planned sale, not a read on the print. Still Labeled $NVDA China 49% verbal guidance to cloud partners — RUMOR Samsung Taylor Fab 2 fully booked for HBM base dies — RUMOR TX/GA governor data-center pause framing — UNVERIFIED Bottom line: Demand for power and interconnect is still being contracted. The cost of that capacity (coupon and operating cost) has moved higher. Talent is being bid for in real time. The next constraint is whether returns clear the new hurdles.
🤖 Made with AI
254
Be more realistic and a vs b options going forward, vs flavor of the day or or rose color glasses. You move money in/out of your investments /cash etc, but your show on TV really doesn't run in parallel that I can see. Challenge guests, don't just give time on TV @cvpayne
1
34
Now @ElonMusk this now is saying that the largest hurdle that AI data centers in space has to overcome, heat dissipation. He claims this is generally easy to solve by simply having the radiators on opposite side from the sun. Now for a proof of concept
🤖 Made with AI
1
1
37