Surrealist5N1K

Nvidia employees are reportedly urging the company to bring Jensen Huang back, with an internal post receiving over 1,600 comments, mostly in support of his return.
The board has yet to announce any decision, leaving the future of Nvidia’s leadership uncertain.
Will Jensen Huang return?
#GateIPOAccessAnthropic

#Gate直通IPO第三期Anthropic Based on public prospectus information, media reports, and the latest product developments, the following analyzes Anthropic and its flagship product Claude across five dimensions: user experience, product competitiveness, business model, commercial prospects, and risks.
1. User experience: A divide between capabilities and limitations
Claude’s user experience has a clear “two-track” character:
· High capability ceiling: Claude Fable 5.2 and Mythos 5.2, released on November 9, 2026, are positioned as the most capable models currently available. They perform strongly in complex software engineering, scientific reasoning, chart comprehension, and robot control.
For example, Fable 5.2 scored 49.1% on the FrontierCode 2.0 coding benchmark, outperforming contemporaneous competitors; it achieved a 64.9% success rate in simulated bimanual robot tests. The new models also demonstrated the ability to coordinate over 1,000 subagents and continuously write an operating system prototype for 8 days.
· Limitations for individual users: Despite the models’ powerful capabilities, individual users—especially free and Pro users—often face usage caps, rate limits, and strict risk controls. Some users have reported that risk controls are relatively stringent in regions such as mainland China, with account bans occurring relatively often.
· The root cause of the experience gap: This divide stems from compute allocation strategies. Given limited compute, the company prioritizes high-value enterprise customers. The prospectus acknowledges that the company “has always operated in a compute-constrained environment” and may adjust usage limits in the future, potentially affecting the stability of the experience for individual users.
2. Product competitiveness: Agents and enterprise workflows.
Anthropic’s core competitive edge has shifted from a simple “chatbot” toward “autonomous agents” and “embedding in enterprise workflows”:
· Claude Code drives usage: Claude Code is seen as a key product defining autonomous AI work. Prospectus data shows that in the third quarter of 2026, the top 5% of enterprise users of Claude Code spent an average of about 20 times as much per month as users who did not use the feature.
This indicates that agents capable of autonomously writing code, running tests, and fixing bugs are a core use case that consumes compute and generates revenue.
· Model lineup and pricing: The current model lineup includes Mythos (including Fable), Opus, Sonnet, and Haiku. The standard pricing for the new models launched in November 2026 is $8 per million input tokens and $40 per million output tokens, 20% lower than the previous generation, while output speed is 30% faster; a higher-performance option is also available in ultra-fast mode.
· Safety and alignment: The company emphasizes “Constitutional AI” and safety research. The Fable version, aimed at general users, comes with comprehensive safeguards, while the Mythos version has fewer restrictions and is available only to vetted organizations. This differentiated strategy meets the public’s safety needs while preserving room for exploration in professional fields.
3. Business model: Dominated by B2B and cloud channelsAnthropic’s business structure is highly dependent on enterprise customers and cloud provider channels, with individual users contributing a relatively small share:
· Revenue mix: In the first half of 2026, usage-based revenue was about $13.064 billion, accounting for around 80% of total revenue; subscription revenue was about $3.222 billion, accounting for around 20%. Subscription revenue is not broken out separately between individual users (Pro/Max) and enterprise plans (Team/Enterprise), but individual users’ contribution is certainly below 20%.
· Dependence on cloud channels: The share of sales through cloud provider marketplaces such as Amazon, Google, and Microsoft continues to rise. In the first half of 2026, cloud provider channels accounted for about 48% of revenue, nearly half.
· Customer structure: As of September 30, 2026, the number of customers contributing more than $100,000 in revenue over the past 12 months had surged from over 950 a year earlier to over 11,900. The top five customers contributed about 25% of revenue in total, keeping dependence on any single customer manageable.
· Customer stickiness: As of September 30, 2026, net revenue retention (NRR) was as high as about 900%. Note, however, that individual subscription users are excluded from this calculation, which primarily reflects expansion among enterprise customers.
4. Commercial prospects: Rapid growth, but costs must be absorbedBased on its financial and operating data, Anthropic is in a high-growth phase, but the quality of its earnings and path to capitalization remain to be seen:
· Revenue growth: Revenue in the first half of 2026 was $16.286 billion, 3.55 times full-year 2025 revenue ($4.585 billion). Revenue for the second quarter of 2026 reached $11.555 billion, up 1,368% year over year.
· Turning point in profitability: The company generated $974 million in operating profit in the second quarter of 2026, with an operating margin of about 8.4%, marking its first profitable quarter at the operating level. This was mainly because revenue growth (1,368%) far exceeded the growth in computing and infrastructure expenses (about 300%), and computing expenses fell from 177% to 48% of revenue.
· Book losses: Despite turning profitable at the operating level, net losses in the first half of 2026 remained as high as $80.748 billion. The bulk—$72.388 billion—came from fair value changes in convertible bonds, derivative liabilities, and SAFEs. These are non-cash accounting expenses and do not directly indicate a deterioration in operating cash flow.
· Valuation and IPO: The company plans to list on Nasdaq around mid-November 2026 under the ticker ANT. Media reports put the valuation discussion range at $1.5 trillion to $2 trillion, compared with a post-money valuation of $965 billion in its May Series H financing.
· Per-employee efficiency: As of September 30, 2026, the company had about 5,300 employees, with revenue per employee over the past 12 months exceeding $6.6 million, demonstrating exceptionally high talent density and operating efficiency.
5. Risks and challenges
Despite its promising prospects, the following risk factors have been cited repeatedly in the prospectus and market analyses:
· Compute supply and allocation: The company acknowledges that customer demand exceeds available compute. With compute perpetually in short supply, balancing resource allocation between high-value enterprise users and individual users could pose a reputational risk.
· Cost pressures: Although computing efficiency improved in the second quarter, some contracted compute capacity has yet to be delivered, and the related costs will be recognized over time. Subsequent cost increases could put pressure on profit margins.
· Competitive landscape: Competitors such as OpenAI are intensifying competition in the enterprise market, and businesses may adopt lower-priced open-weight models. Anthropic’s revenue growth in the third quarter of 2026 (45% quarter over quarter) was already below the second quarter’s rate (144% quarter over quarter), suggesting growth may be starting to slow at the margin.
· Conversion of individual users: Compute consumed by free users is included under “sales and marketing” expenses. The company hopes to convert them into paying customers, but the prospectus also acknowledges that conversion may not succeed and that the free tier could cannibalize paid subscriptions.
· Accounting and valuation volatility: The massive reported net losses are primarily due to fair value changes in financial instruments. Investors should distinguish operating profitability from accounting losses. In addition, IPO valuation is influenced by market sentiment and may be subject to adjustment.
Summary
Anthropic currently demonstrates extremely strong enterprise commercialization capabilities. Products such as Claude Code are deeply embedded in enterprise workflows, driving exponential revenue growth and a turn to operating profitability. Its core strengths lie in the vertical integration of model capabilities, safety alignment, and enterprise channels. However, its commercial prospects remain constrained by compute supply bottlenecks, high infrastructure costs, and intense market competition.
For individual users, although the models are top-tier in capability, they may remain a lower priority in the product lineup for the long term. If the company can continue to grow revenue faster than costs, its long-term value will be further validated; conversely, if compute costs spiral out of control or enterprise customer growth slows, its high valuation will face a test.

#黄仁勋辞职 This is absolutely not the time to go all in blindly; wait for 3 key signals to materialize before taking action.
📉 NVIDIA’s current core situation (risks far outweigh opportunities)
1. Jensen Huang’s sudden departure directly removes NVIDIA’s most important “industry coordination hub” of the past 10 years. He personally secured end-to-end ties spanning TSMC’s capacity, the HBM supply chain, global AI clients, and the developer ecosystem. This is the core moat that has enabled NVIDIA to outpace AMD and Broadcom, and it cannot be readily replaced by an existing product roadmap.
2. The market has already voted with its feet: since its October all-time high, NVIDIA’s market cap has shed about $1.7 trillion, a cumulative decline of 31%. After the news broke, its biggest premarket drop hit 12%, while short sellers’ paper gains for the day alone exceeded $3.8 billion.
3. The current interim CEO, Kress, has a deep finance background and can stabilize short-term earnings and cash flow, but she lacks Jensen Huang’s ability to break new ground in the industry. The market has no idea whether she will cut high-risk frontier R&D or scale back ecosystem investment. This is the biggest hidden uncertainty.
🎯 My trading plan (execute in three steps)
1. Stay on the sidelines with no position. Don’t buy the dip at all or try to catch a falling knife.
2. First observation window: the Q3 earnings report on November 17
Focus on two details: first, whether the revenue outlook is lowered due to the management change; second, whether the interim CEO is vague about the supply chain and major clients on the earnings call. If the earnings greatly exceed expectations and the outlook is not cut, that is only the first safety cushion—it is still nowhere near the point where it makes sense to take a heavy position.
3. Second observation window: the appointment of a permanent CEO
This is the decisive signal. If a senior internal technical executive takes over and can continue Jensen Huang’s product roadmap, I can add back 50% of the position I previously reduced. If an outside professional manager parachutes in, or someone with a purely financial background takes the helm, I’ll abandon the plan entirely. That would mean NVIDIA’s growth story has fundamentally changed.
4. Contingency plan for an extreme scenario: If the share price later plunges in a panic to a drawdown of more than 45%, and there is no concrete evidence of core customer losses or supply-chain disruptions, then I’ll use 15% of my total portfolio for a small, contrarian position, with absolutely no excessive leverage.
⚠️ Two traps to avoid
• Don’t cling blindly to the argument that “the long-term AI sector outlook is positive.” 80% of NVIDIA’s current premium is for Jensen Huang’s command of the ecosystem, not simply for its chips. If that thesis breaks, the valuation framework will be completely reworked.
• Don’t assume a 10% short-term drop makes this a bargain. The entire AI sector is under pressure now, Nasdaq futures are down more than 1%, and competitors like AMD and Broadcom are still taking advantage of the situation to grab market share. The downward momentum is much stronger than you might think.$NVDA ‌

#黄仁勋辞职 Jensen Huang resigns, marking a pivotal turning point for Nvidia: What’s next for the AI giant?
A sudden leadership change has thrust Nvidia into the global tech spotlight. On November 12, 2026, multiple media outlets reported that Jensen Huang had resigned as Nvidia’s president, CEO, and director, effective immediately. Current CFO Colette Kress has become interim CEO, and the board will search for a permanent successor from inside and outside the company.
Why has the market reacted so strongly?
After news of Huang’s departure was announced, Nvidia shares fell about 10% in U.S. premarket trading. Note that this is an immediate premarket move, not the final closing performance for the day.
The core concern is the uncertainty brought by a sudden change in leadership.
From a business perspective, when a company’s technological direction, customer relationships, and public image are closely tied to its founder, a leadership change prompts investors to reconsider several questions: Can the new management maintain decision-making efficiency? Will key customers and partners continue their support? Can product and business plans proceed on their original timelines? Does the company have a sufficiently mature succession process? Subsequent business performance will have to answer these questions.
Short-term share prices reflect uncertainty, while long-term value still needs to be proven through business performance.
Who is interim successor Colette Kress?
According to Nvidia’s official biography, Colette Kress joined the company in 2013 as executive vice president and CFO. Before that, she worked at Cisco, Microsoft, and Texas Instruments, gaining extensive experience in financial management in the tech industry. Having an executive familiar with the company’s finances and operations take over on an interim basis may help the company navigate the transition from the standpoint of management continuity. But the interim arrangement is only the first step. The choice of a permanent successor, how technical and business decisions are divided, and how the board oversees management will all shape the market’s view of Nvidia’s future.
What does this mean for the AI industry?
First, the market may reassess tech companies’ reliance on their founders.
A founder’s influence can help a company grow quickly. But when key decisions, customer relationships, and public trust are highly concentrated in one person, the impact of a leadership change is amplified. Investors may now focus more on whether the company’s team, systems, and execution capabilities can continue to support growth after the founder’s departure.
Second, corporate governance may become even more important.
For companies at the center of global technological competition, investors look beyond revenue and profits to information security, conflict-of-interest management, internal oversight, and succession planning. These systems rarely make headlines in normal times, but when a crisis occurs, they often determine whether a company can steady itself.
Third, leadership changes and demand in the AI industry should be assessed separately.
Huang’s departure alone does not directly prove that AI demand is weakening, nor does it prove that competitors have already won orders. Assessing the industry’s direction still requires looking at customer spending, product deliveries, profitability, and upcoming financial reports. It is too soon to equate a single leadership crisis with a reversal in the entire industry’s trajectory.
What should we really watch next?
Rather than repeatedly chasing unverified rumors, it is more worthwhile to watch three signals:
Investigation progress.
Will the company disclose new facts, and will it need to adjust its internal systems?
Succession plans.
When will a permanent CEO be appointed, and how will the core management team divide responsibilities?
Business performance.
Will there be material changes in orders, deliveries, profitability, or customer partnerships?
For Nvidia, this is a test of leadership, corporate governance, and market trust.
After Huang leaves his management role, whether the company can continue to maintain technological innovation and execution efficiency will gradually become clear through subsequent decisions and business results.
Capital markets can reprice a company within minutes, but its long-term competitiveness ultimately has to be proven through its products, team, and customers.$NVDA ‌

On November 10, 2026, global risk assets presented a picture that left many traders puzzled. The three major U.S. stock indexes all rebounded strongly, with the Dow closing up 2.15%, the S&P 500 rising 2.35%, and the Nasdaq gaining 2.74%. Spot gold also climbed, closing near $4,140.94 per ounce. Meanwhile, Gate market data showed Brent Crude Oil at $90.86 per barrel, down 6.02% on the day, and WTI at $85.61 per barrel, a decline of 5.03%. The sharp drop in oil prices followed a clear headline: r$NVDA $MU $AMD $SPCX

#Polymarket年化收入突破18亿美元 $1.8 billion in revenue is real, but the “takeoff” may not land on Polymarket
The prediction market sector is definitely taking off, but Polymarket isn’t necessarily the sure-fire beneficiary.
I. First, separate the two numbers
$1.8 billion is annualized revenue, or about $150 million per month; $14.5 billion is the US site’s nominal trading volume in October. A rough calculation puts the monetization rate around 1%—but note that the figures use different scopes: revenue is global, while the $14.5 billion is only for the US site, so this is just an order-of-magnitude reference.
More importantly, there’s the term “nominal trading volume.” Each prediction market contract ultimately settles at $1, so a trade bought at $0.30 and sold at $0.70 generates nominal volume twice. That means $14.5 billion isn’t $14.5 billion in capital; the actual capital tied up is an order of magnitude smaller.
What does this mean? The industry makes money through scale, not commissions. The apparent boom in trading volume translates into less financial leverage than you might think.
II. Where growth comes from: the new-market launch boost plus seasonal traffic, not structural growth
The US site only launched in May this year. In October, nominal volume reached $14.5 billion, up more than threefold from just $4 billion in June. The drivers are clear: the FIFA World Cup in the summer, followed by the start of the NFL season in September.
In other words, this was a one-off convergence of “launching in a new market” and “seasonal traffic.” 2027 has no World Cup and no major election, making it a typical “event lull” year. Whether financial predictions like “CPI, earnings, and BTC prices” can carry the business then will be the real test.
III. The “takeoff” may not benefit Polymarket: market share has already flipped
Here’s what’s easiest to overlook: at the end of 2024, Polymarket still accounted for over 90% of monthly trading volume. Now the tables have turned.
In August 2026, Kalshi did $37.17 billion in a single month, while Polymarket’s US and international sites combined did $8.16 billion; from mid-September to mid-October, the shares were roughly 76% for Kalshi, 12% for Polymarket’s US site, and 5% for its international site.
So Polymarket’s $1.8 billion came amid a backdrop of “losing the lead in market share”—its growth came from the overall pie getting bigger, not from its share increasing. These two things have completely different implications for valuation.
IV. The real bottleneck isn’t demand, it’s regulation
Polymarket’s US site operates on QCX, the CFTC-designated contract market it bought for $112 million in July 2025. It only opened to US users in December 2025, starting with sports.
But that’s also where the trouble lies: the New York attorney general has already sued it for illegally operating a gambling business, and Nevada’s gaming regulator has taken direct enforcement action. The CFTC’s draft regulatory rules in June this year left a relatively permissive opening for election-related political markets while drawing a red line around sports-event betting.
In other words, the industry is now facing a legal battle between “federal derivatives regulation” and “state gambling laws.” This uncertainty matters more than any growth figures.
V. Which area has the most potential? My answer is sports, not finance
This may go against intuition, but I’m increasingly convinced.
Sports has the most potential because it’s a “legal alternative to gambling,” with the largest TAM and the highest repeat usage. Retail users also face the smallest information disadvantage—everyone knows little about sports, so they’re still willing to bet. The World Cup and the NFL are just the beginning.
I don’t think financial events (CPI, FOMC, earnings) can become a retail traffic pool, because the odds are too “efficient”: informed professional players will keep winning money from retail users, who won’t come back after a couple of tries. These markets are more like hedging tools and information derivatives. They’ll be big, but they won’t be lively.
Politics, meanwhile, comes around once every four years and has the most settlement disputes.
VI. What I would personally trade
I’d only trade financial events, and only those with clear settlement rules and enough depth: CPI, FOMC decisions, and earnings from large companies. I wouldn’t touch sports: fees, an information disadvantage, and emotional trading—all three are in play. I wouldn’t touch politics either, given the settlement disputes and regulatory uncertainty.
But there’s one reality to accept first: research shows that 70% of top profitable wallets are bots. In other words, if you enter the market, you’re effectively “the one providing liquidity to market-making bots.” Think that through before talking about the experience.
The $1.8 billion proves that “event trading” is real demand, but it doesn’t prove Polymarket is the winner. The moat in this sector isn’t “events”—it’s regulatory access, market-making depth, and distribution channels. Kalshi has secured the first two, plus Robinhood’s distribution; Polymarket has crypto-native liquidity, but it still hasn’t cleared the regulatory hurdle.
Would you start with CPI or sports? Share your first trade in the comments.

#美国10月CPI全面超预期 When data and price conflict, I trust price first, but only use it to size my position
If PPI continues to come in above expectations and BTC does not break below its previous low, I will trust price more. But there is one condition: I only use price to decide whether I can act now; direction and potential upside or downside still depend on macro.
I. First, understand the marginal information in this CPI report
The market has already priced in an 88.5% probability of a 25-basis-point rate hike in December, while the probability of a 50-basis-point hike is only 11.5%; by January next year, the probability of cumulative rate hikes totaling 50 basis points is 59.2%.
In other words, most of the impact of the phrase “across-the-board upside surprise” has already been absorbed by prices. What was actually shaken out was positioning, not valuations: $657 million in BTC liquidations in one day, including $540 million in long positions, while open interest actually rose 5% to 341.5 billion. This is a typical “positioning event,” not a “valuation event”: leverage has not been flushed out, so the decline has been neither clean nor decisive.
So the answer to “how much more pressure is there to withstand?” does not lie in how much October CPI exceeded expectations, but in whether “another hike in January next year” will be put on the table. That is the source of the second round of pressure.
II. Why I don’t set the two against each other
Macro tells you where the ceiling is: the rate path is not over, the 10-year yield is above 5%, and the dollar is still strengthening. In this environment, BTC has no independent rally; a rebound is just a rebound, not a reversal.
Price tells you whether you can act now: if bad news fails to push prices down, it means marginal sellers are drying up. This is information about the supply of coins, reflecting who is still selling and who has already finished selling, rather than fundamental information.
The two are actually answering different questions. The real danger is using them in reverse: seeing that “it can’t fall any further” and inferring that “rate hikes are about to end.” That substitutes market positioning for macro judgment, and usually comes at a cost.
III. My actual trading plan (BTC)
Going long requires all three conditions to be met simultaneously; if even one is missing, I won’t act:
First, on the day PPI exceeds expectations, BTC’s daily close must hold above 72,787 (the previous low in this move), and the 4-hour lows must be rising.
Second, declining volume on the way down and no volume chasing on the rebound would indicate that sellers are truly exhausted, rather than being artificially held back by capital.
Third, the funding rate must turn negative or approach zero, with longs no longer crowded.
If all three are met, I’ll scale in: first tranche at 73,000 to 73,500, second tranche at 71,800 to 72,300, with a stop-loss at 71,500 (about 1.5% below the previous low), and an initial target of the structural recovery zone at 76,000 to 77,000. Leverage will not exceed 2x, and total position size will not exceed 30%.
If the daily close breaks below 72,787, then macro has won, and the target moves down to 70,000 or even 68,000. I won’t catch a falling knife; I’ll wait for a second sign of stabilization near 68,000.
IV. What about gold and U.S. stocks?
Gold: 4,000 is a psychological level, and 3,950 to 3,980 is the previous-high platform. If PPI surprises to the upside again and real yields continue to rise, gold will remain under short-term pressure; but buying around 4,020 shows that the long-term thesis remains intact. My approach is short-term bearish, long-term bullish: only a break below 4,000 coupled with continued dollar strength would count as a real warning.
U.S. stocks: The Nasdaq 100 closed at 27,740; a forward P/E of 18x is a valuation floor, not an earnings floor. A break below 27,421 (the November 10 low) points to 26,500. But the real risk for U.S. stocks is not this CPI report; it is downward revisions to 2027 earnings expectations.
Data tells you whether you should be afraid; price tells you whether you should act now.
Most of the marginal information in this CPI report has already been priced in. The real variable from here is whether PPI will put “another hike in January” on the table. So I won’t predict direction; I’ll wait for the signal: when PPI comes out, price will give the answer.
Where do you place the “previous low” yourself: 72,787 or 72,000? Let’s compare in the comments. #WCTCS9

#美国10月CPI全面超预期 #每周来晒 The Fed doesn’t need to “turn hawkish”; it only needs to maintain a hawkish stance.
I. There are three levels to “turning hawkish,” and the market has only priced in the first
First level: a 25-basis-point hike in December. The probability is 88.5%, already fully priced in. This isn’t turning hawkish; it’s confirmation.
Second level: a one-off 50-basis-point hike in December. The probability jumped from 0% before the release to 11.5%. This is a real “hawkish turn,” and it’s also the line Waller drew himself just hours before the data: a core monthly increase of 0.3% or higher should be considered. The result was exactly 0.3%, right on the threshold.
Third level: changing the path. The probability of cumulative hikes totaling 50 basis points by January next year has reached 59.2%. This is what the market is really repricing. Whether the neutral rate moves higher in the dot plot matters more than the size of this hike.
My view: December will most likely still bring a 25-basis-point hike. The reason isn’t just officials’ comments; there’s also a technical detail many people have overlooked—this CPI report had no year-over-year data because no price data was collected during the October 2025 government shutdown, leaving no baseline.
Without the year-over-year figure as an anchor, it’s difficult for the committee to characterize inflation with a “one-off 50-basis-point hike.” It’s not that they don’t want to; they don’t have enough evidence. So the missing data actually makes a 50-basis-point hike less likely.
II. Where are the limits of hawkishness? Watch three things
The monthly change in core PCE, in the final report before the December meeting.
Whether there will be another hike in January next year. This is a question of the path, and the real source of pressure on BTC.
Whether the statement’s definitions of “additional” and “restrictive” have changed. Changes in wording often come before changes in the numbers.
III. This time, BTC fell because of leverage, not valuation
Price-wise: it fell below 73,000, with an intraday low of 72,787; ETH fell 4.75%, SOL fell 4.9%, and ZEC fell nearly 13%.
The structure is more worth watching: 24-hour liquidations totaled $657 million, including $540 million in long positions; open interest, meanwhile, rose 5% to 341.5 billion; the total crypto market cap fell 2.93%, while trading volume rose 15.5%.
Prices down, trading volume up, open interest up—that’s a classic “longs being forcibly liquidated as new shorts enter” structure. The Fear and Greed Index is at 24, and BTC dominance is 60.4%, showing that money is taking shelter in the hardest assets rather than leaving the market.
So the characterization is clear: this is liquidation-driven, not allocation-driven.
IV. Why is BTC so sensitive to CPI? Three transmission channels
First is the discount rate. The 10-year yield broke 5%, and the 2-year yield is at 4.71%. The numerator in the valuation of long-duration assets hasn’t changed, but the denominator has become more expensive; BTC is among the “longest-duration” assets in this cycle.
Second is the dollar. The dollar index rose 0.56%. When offshore dollar liquidity tightens, BTC’s marginal buyers—the leveraged longs—are the first to pull back. BTC’s pricing power has always been in the hands of offshore dollars, not Wall Street.
Third is leverage. When longs are crowded, data becomes the trigger. It’s not that CPI is so frightening; it’s that positions are too full.
V. How to distinguish deleveraging from a loss of conviction
What happens next is completely different for these two, but they look very similar on the charts.
Signs of deleveraging: open interest drops rapidly, funding rates turn negative, the spot premium recovers, and liquidations are concentrated within one to two days. This usually ends within a week.
Signs of a loss of conviction: open interest remains stable, but spot outflows continue, ETFs see consecutive redemptions, and long-term holders begin reducing their on-chain positions. This takes months.
What we’re seeing now is the early stage of the first scenario—note, the “early” stage—because open interest is still rising, which means leverage hasn’t been fully cleared out. So don’t rush to call a bottom.
VI. My plan
I don’t predict direction; I only act when conditions are met.
Only if all three conditions are met—a daily close above 72,787, declining volume on the way down, and negative funding rates—will I scale into longs, with a stop at 71,500 and an initial target of 76,000 to 77,000.
If the daily close falls below 72,787, I’ll look at 70,000 or even 68,000. I won’t try to catch a falling knife.
Before the December 9 rate decision, total exposure will stay below 30%, and leverage below 2x.
The Fed doesn’t need to “turn hawkish”; it just needs not to turn dovish. For BTC, the real bottom isn’t “CPI peaking”; it’s “leverage being cleared out.”
Do you think this cycle is about deleveraging or a loss of conviction? Share the reasons behind your view in the comments. $BTC

#Gate股票现货全球前三 After the pullback, my tech stock watchlist
Gate’s stock spot offering made the global top three on RootData (score: 70.1, 24-hour trading volume: $65.88 million, share: 4.77%), so the range of assets is definitely broad enough. But the ranking is just a ticket in; what follows is what I’m really watching.
I. First, my market view
The first valuation reset is over; the earnings reset hasn’t happened yet.
The Nasdaq 100 closed at 27,740, and its forward PE returned to 18x, the same level as before ChatGPT launched in November 2022; its PE-TTM is 26.4x, at the 1.35th percentile of the past year. But S&P earnings growth in Q3 was 33% to 35%, likely the peak of this cycle, while the 2027 consensus has already fallen to 18%, and hyperscalers’ capex growth is set to drop from 96% to 44%. Valuations have already adjusted; earnings are only just beginning to.
Second, the headwind is interest rates, not the AI narrative.
The 10-year yield broke 5%, the 2-year yield is at 4.71%, the dollar index is strengthening, and an interest rate hike of 25 basis points in December is priced in at 88.5%. In this environment, every rebound in long-duration assets lacks one prerequisite: “the denominator stops getting more expensive.” The AI story hasn’t gone bad; the discount rate has.
Third, a reversal requires two signals to appear at the same time.
One is Nvidia’s guidance on November 17; the other is the Fed’s wording on December 9. If either one turns positive, the rebound can qualify as a reversal; if neither does, it’s still a “relief rally after bad news is priced in,” followed by more grinding lower.
II. In volatile periods, I choose depth over breadth
Breadth answers “can I buy it?”; depth answers “am I paying too much, and can I sell it?” During sharp swings in tech stocks, what you really pay is slippage: with thin order books, spreads can eat up 1% to 2%, and a round trip costs as much as one stop-loss.
So in the list below, I’ve deliberately included only names with “depth”—not a single obscure stock.
III. Watchlist (five groups, each with a trigger)
Group 1: Event-driven, watch only
Nvidia, $168.89, PE 21x, forward PE around 14x. Earnings after the close on November 17, with market expectations of around $109 billion in revenue and EPS of 2.47; closed-door questioning by three House committees on November 18; GTC from November 30 to December 3. The risk it faces is governance, not demand. The trigger is price: below $145 is the first tranche, while $120 to $130 is the extreme “CEO gets fired” scenario. I won’t buy a single share before the events play out.
AMD, $515.21, PE 110x. Q3 revenue was $13.4 billion, up 45% year over year; data center revenue was $8.5 billion, up 97% year over year; and Q4 guidance was $17 billion—the results are genuinely strong. But a 110x multiple already fully prices in 2027, and the stock rose 11% and 13% on November 6 and 9, respectively, making it a classic sentiment-driven stock. Trade it only for swings, and take profits quickly.
Group 2: Certainty, add on pullbacks
TSMC, $396.48, PE 25.7x. The most diversified customer base among “AI tax collectors,” without the burden of relying on a single customer or CEO. This is the best value-for-money stock on my list; I’ll add in the $370 to $380 range.
Broadcom, $331.98, PE 41x. ASICs plus networking, with long order visibility, but the valuation isn’t cheap; suitable only for dollar-cost averaging, not chasing.
Group 3: Valuation discount, my top pick
Google, $287.38, PE 15x. The only one of the Magnificent Seven back at its 2022 valuation. Q3 revenue rose 25% year over year, while net income fell 25% year over year—the profits have been eaten up by depreciation, but that also means its profit upside will be the greatest once it’s done absorbing 2027 capex.
Microsoft, $533.16, PE 27.8x. It lowered 2026 capex from $175 billion to $170 billion. In a market where “whoever spends the most gets punished,” restraint itself is good news.
Group 4: Defensive, to reduce volatility
Apple, $348.51, PE 38x. No capex exposure and the cleanest cash flow make it the cleanest safe haven this cycle, but at 38x it can only defend, not attack.
An even harder asset class is power and gas turbines—the end point of computing power is electricity, data center power consumption is one of the most rigid components of capex, and it isn’t exposed to the valuation reset in AI chips.
Group 5: Watch only
Meta, $641.39, PE 21.8x. Q3 free cash flow turned negative at $2.12 billion, and it still plans to issue $35 billion in equity financing. It’s the best stress-test case for the “AI capital cycle”—I’ll consider it once it proves it can turn capex into free cash flow.
IV. How I’ll execute
I won’t buy everything at once. The order is: start with Google and Microsoft, then wait for the November 17–18 events to play out before adding Nvidia; add TSMC on a pullback, and trade AMD only for swings.
No single position will exceed 15% of my total portfolio; during the event period from November 17 to December 9, total exposure will stay below 50%.
Stop-losses are based on earnings expectations, not price. Once 2027 earnings expectations start to be systematically revised downward, I won’t touch the stock, no matter how cheap the valuation gets.
This list isn’t about “what to buy”; it’s about “when to buy, how much to buy, and when not to buy.” In volatile periods, half the gains come from choosing the right stocks, and half from not going all-in at the wrong time.
Which ones are on your list? Let’s talk in the comments.

#OpenAI第三季度营收升至100亿美元 Bullish on AI, but retail investors can’t buy “model companies”
I. First, read OpenAI’s report card in full
Q3 revenue was $10 billion, up nearly 50% quarter over quarter, with an annualized run rate now at about $50 billion; the previous quarter was $6.7 billion.
But there’s another side to the same report card: estimated losses of about $14 billion in 2026, and cumulative investment projected to reach about $115 billion by 2029. Altman himself has said that continuing to grow at this rate will require “trillions” in infrastructure investment.
The valuation tells an even clearer story: in March, it raised $122 billion at an $852 billion post-money valuation, setting a record for a single funding round by a global startup; the target valuation for the current round is $1.25 trillion, but only $11 billion has come in during the first tranche, with both the size and valuation expectations shrinking from the previous round.
So the company’s profile is this: revenue is growing exponentially, losses are expanding in proportion, and the faster it grows, the more it depends on capital markets to keep funding it. It is the world’s most expensive “work in progress.”
II. The key question: Can you buy it?
OpenAI is not publicly listed. The March round was the first to open to individual investors through bank channels, raising more than $3 billion from individuals, while also being included in several ARK ETFs.
In other words, ordinary investors have only two ways to get exposure to it: qualified-investor shares through bank channels, or buying ETFs that hold it or shares of its corporate shareholders.
So “model companies or the supply chain” is really a false choice for most people—you can only pick from the supply chain, or use the shell of a public company to buy exposure to a model company.
III. What you should really bet on is where profits migrate along the chain
The reality today is that hardware takes most of the profits. The AI infrastructure sector accounts for half of S&P 500 earnings growth, while Nvidia is still growing revenue at 100%; profits decline at each step along the “chips—computing infrastructure—large models—applications” chain.
But institutions’ medium-term outlook points in the opposite direction: hardware’s share of profits will shrink, while revenue for models and applications could grow 26.9-fold from 2026 to 2030 (versus 12.5-fold for cloud services), driving nearly two-thirds of incremental earnings.
These two statements don’t contradict each other; they describe different points in time.
Selling shovels is “the best business now”; models are “the best business of the future.” They aren’t opposing camps, but upstream and downstream parts of the same chain—the real bet is on which end profits will migrate to.
IV. My allocation strategy: 70% supply chain, 30% model exposure
Within the supply chain, I only buy companies with contract-backed demand and reasonable valuations:
Nvidia at $168.89 and 21x PE has the longest order visibility, but its November 17 earnings report and November 18 congressional hearing are two simultaneous events, so it’s not suitable for a full-position bet.
TSMC at $396.48 and 25.7x PE has the most diversified customer base, so it doesn’t have to bear the risk of relying on a single customer.
Oracle is a “long-dated payoff” play—its $300 billion in contracts will take effect in stages under 5-to-6-year capacity agreements, with most of it not coming due until 2028. Buying it means buying patience.
AMD has 6 gigawatts of computing capacity agreements, but its 110x PE already prices in the full expectations, so it’s only for swing trades.
I don’t buy model exposure directly; I use the shells of public companies:
Microsoft at $533.16 and 27.8x PE is one of OpenAI’s largest shareholders, while also being a platform in its own right.
Amazon at $270.87 and 21.7x PE has invested $50 billion in OpenAI and also holds an equity stake in Anthropic.
Google at $287.38 and 15x PE has a stake in Anthropic as well as its own models, and is also the cheapest of the Magnificent Seven.
What these three have in common is that their cash flows can cover AI investment, so they don’t need to rely on fundraising to stay alive. Buying them is like buying a “public company shell + model company option”—the only viable path for retail investors.
V. When will this answer change?
Anthropic’s IPO. Its Q3 revenue was 70% higher than OpenAI’s, with an IPO valuation cap of about $1.41 trillion; Amazon’s $18 billion investment could swell to $270 billion, while Google’s $33.3 billion investment could reach $139 billion.
Once it goes public, model companies will have a public-market valuation benchmark for the first time. At that point, “models or the supply chain” will finally become a choice you can actually make, rather than one you can only watch.
The easiest mistake to make when bullish on AI is to equate “being optimistic about the industry” with “buying the purest play.” The purest plays are often unavailable, and the ones you can buy are often not the purest.
As profits migrate along the chain, upside potential matters more than purity.
Would you buy private shares for “purity,” or the supply chain for “upside potential”? Let’s discuss in the comments#Gate股票畅聊季

#美股AI概念股普跌 The AI industry is still growing, but the market is starting to discount expectations of “unlimited investment.”
On November 5, the Nasdaq fell 5.52%, and the Philadelphia Semiconductor Index fell 12.25%. Chip stocks fell the hardest, while cloud providers like Microsoft and Amazon saw much smaller declines. This shows that investors are reassessing companies, not that AI demand has suddenly disappeared.
Why did the market fall this time?
Policy expectations shifted: After the midterm elections, data center approvals and AI regulation may tighten.
Returns are being questioned: Free cash flow at Google, Meta, and others has turned negative, putting the problems of heavy investment and slow returns in the spotlight.
Interest rates are weighing on valuations: Rising Treasury yields are hitting high-valuation tech stocks first.
Crowded trades are unwinding: Leveraged ETFs and quantitative funds cut positions in unison, amplifying the declines.
What signs suggest the trade is really cooling?
Earnings expectations have not been significantly revised down; stock price declines are mainly driven by P/E multiple contraction, indicating that valuations are being repriced.
Companies with higher capital expenditures have fallen harder, as the market starts scrutinizing returns on investment.
Pure-play concept stocks and highly leveraged computing-power rental companies face more pronounced shakeout pressure.
Bernstein believes the long-term thesis remains intact, but a full recovery will take time.
Looking ahead, markets will remain volatile in the short term. For the market to truly enter a cooling phase, watch whether data center construction actually slows, cloud providers cut spending, and AI revenue continues to materialize. For now, this looks more like a shift from “telling stories” to “checking results,” not a bubble bursting like in 2000. #WCTCS9

#直通IPO第三期Anthropic I’m most bullish on Anthropic selling Claude into enterprise workflows. Revenue is primarily usage-based, and as customers move from trials to long-term purchases, the growth potential is far greater than simply selling models.
Most promising growth drivers
Existing customers are using more and more: As of the end of September, there were 11,900 customers paying more than $100,000 annually, with net revenue retention of around 900%.
Agents drive usage: Claude Code and Cowork can continuously complete tasks, and complex tasks consume more compute, increasing revenue.
Low penetration of enterprise budgets: Top enterprises spend an average of less than 0.1% of their total expenses with Anthropic, leaving significant room to expand.
Improving inference efficiency: In the first half of 2026, compute and infrastructure costs accounted for about 55% of revenue, down from 201% in the same period last year.
Offering plan
Anthropic plans to offer shares at $70–78 each, with a fully diluted valuation of up to about $1.405 trillion. Pricing is set for November 18, trading begins on November 19, under the ticker “ANT.”
I will participate in this offering, but only as a watch position. The high valuation already prices in very high growth expectations; going forward, revenue growth, operating profit, and cash flow still need to deliver.
The main risks are also clear: Long-term compute purchase commitments amount to about $900.9 billion, with nearly half due by the end of 2030; about 48% of revenue comes through cloud marketplace channels, which also incur channel costs.
So my view is: The business model is attractive, but this is a high-expectation, high-volatility asset, so it is more appropriate to keep the position within an affordable range.

#直通IPO第三期Anthropic $69.3 Billion IPO: AI Bubble or the Next Decade’s Major Bull Run?
Anthropic’s IPO roadshow may be the most dramatic moment in the 2026 capital markets.
The day before, the market was discussing an offering plan seeking to raise about $69.3 billion at a valuation approaching $2 trillion; the next day, the filing was withdrawn and replaced with a more restrained offering range of $70–$78 per share, with the public offering’s fundraising cap adjusted to about $64.4 billion. Combined with a $26 billion private placement, the total could reach about $90.4 billion. Valuation expectations also fell from $2 trillion to $1.1 trillion–$1.4 trillion. This was not simply a pricing mistake, but a market repricing of AI assets.
Anthropic’s prospectus is half myth, half bill.
The myth is dazzling: as of the end of October 2026, annualized revenue was about $81 billion; third-quarter revenue was $16.96 billion, up more than 11-fold year over year; and the number of enterprise customers paying more than $100,000 annually reached 11,900. From $41 million in 2023 to $81 billion today, this growth curve has almost outpaced Google’s and Meta’s pre-IPO growth.
But the bill is just as staggering: the company has signed commitments to purchase about $900.9 billion in computing capacity and infrastructure, with about $70.7 billion due in the next 12 months. In the third quarter, the company’s operating profit turned into a $567 million loss from a $974 million profit, as it temporarily procured computing capacity at higher prices.
The structure of the participants is more complex still: seven giants—Google, Nvidia, AMD, Broadcom, Micron, Microsoft, and TSMC—have subscribed to a combined $26 billion in private placements. They are customers and shareholders of Anthropic, as well as suppliers of computing capacity and chips. This creates a closed loop: Anthropic trades equity for computing capacity, computing suppliers trade orders for equity, and companies across the industry chain boost one another.
Is this a bubble or a new cycle? More accurately, it may be both.
The bubble side is that valuations are still based on the assumption of “continued ultra-high growth in the future.”
If AI model prices keep falling, open-source models catch up, and corporate budgets tighten, trillion-dollar valuations will look burdensome. Especially as sequential revenue growth falls from 144% to 47%, the market has already begun recalculating its margin of safety.
The new-cycle side is that AI is no longer just a concept; it is entering enterprise workflows.
Claude Code, Claude API, enterprise knowledge bases, and automated workflows are moving from “trials” to “paid use.” When a company’s annualized revenue exceeds $80 billion, it can no longer be simply classified as a “story stock.”
The real question is not whether AI has value, but where that value will accrue.
Over the past two years, capital markets rewarded “more powerful models”; going forward, the market will care more about “who can make more money.” Can computing costs come down? Can customer retention be sustained? Can the application layer generate stable cash flow? These are the factors that will determine whether AI enters a major bull run or faces a valuation shakeout.
For ordinary investors, a more practical approach is to treat it as a barometer of the AI cycle: if Anthropic ultimately prices its shares conservatively and the overallotment option is exercised, that would indicate continued confidence in leading AI assets, and sectors such as computing capacity, chips, and cloud services could continue to benefit; if the offering price is lowered and subscriptions are tepid, it could mean the AI sector is entering a period of stress testing, making it more dangerous to chase concept stocks; if the stock fluctuates sharply after listing, ordinary investors should be even more careful to avoid emotional buying, especially bets made with leverage or concentrated in a single asset.
For most investors, rather than chasing the halo of the “largest IPO in history,” it is better to focus on AI applications that can genuinely generate cash flow, reduce costs, and improve efficiency.
The next wave of AI opportunities may not lie in the model companies with the highest valuations, but in the application layer that turns AI into a productivity tool.
Anthropic’s IPO is not just a pricing experiment for the capital markets, but also a watershed moment as the AI industry moves from “telling stories” to “reporting earnings.” Trillion-dollar valuations can be exciting, but what is truly worth watching is always cash flow, customer spending, and the path to profitability. The trend will not disappear, but the people most likely to lose money in it are often those who see only the valuation and not the bill.
👇What do you think? Welcome to share your thoughts in the comments～$ANTHROPIC ‌

#Gate全球合规版图扩展 Compliance capabilities are becoming as critical a barrier as liquidity.
Why compliance is becoming increasingly important
Institutional funds only enter licensed platforms: without U.S. licenses and auditable custody, compliance departments won't approve them.
Offshore arbitrage opportunities are shrinking: the EU's MiCA is now fully in effect, and the Middle East and Southeast Asia are also tightening regulations.
Smaller platforms are being squeezed out first: AscendEX, for example, halted services over licensing and compliance issues.
How to view Gate's footprint
Broad coverage: 81 jurisdictions spanning crypto trading, custody, payments, and other areas.
Its strength lies in securing high-value licenses in several key markets, with separate licenses in each jurisdiction. This carries more weight than simply having a large number of licenses.
High-value licenses: It holds Malta MiCA and Dubai VARA licenses, as well as Japan FSA and Australia AUSTRAC registrations.
Separate licenses in each region: In Europe, it has both a Lithuania registration and a Malta MiCA license, providing greater resilience to regulatory risks.
Broad coverage of U.S. state licenses: Gate US has obtained about 30 state money transmitter licenses, covering 40 jurisdictions.
High reserve transparency: Its total reserve ratio is about 127%, disclosed using zero-knowledge proofs; RootData ranks it second globally for transparency.
These licenses can help Gate enter more markets and make it easier to attract institutional funds.