Stockgecko

    24/7 Futures Markets by Open Interest

    Open Interest—24h Volume$0.00
    Avg Funding (1D)—
    Top Gainers
    • —

    AI Is the Most Deflationary Force in Human History. You Will Never Feel It.

    The cost of intelligence has fallen 99.9% in 26 months. Central banks exist to prevent prices from falling. The collision between these two forces will define the next decade of financial markets.

    Something unprecedented is happening in the global economy, and the framework most people are using to understand it is wrong. AI is pushing the cost of cognitive labor toward zero at a speed that has no historical parallel. The cost of producing one million tokens at GPT-4 intelligence has collapsed from $60 in November 2023 to $0.05 in January 2026 — a 99.9% decline in 26 months. Chinese models are now offering comparable intelligence at $0.50 per million tokens. Open-source models running locally are pushing marginal cost toward zero. This is not a pricing war. It is the fastest commodity deflation in the history of technology. Faster than Moore's Law drove down the cost of compute. Faster than fiber optics drove down the cost of bandwidth. Faster than solar drove down the cost of energy. The cost of intelligence — the most valuable input in the modern economy — is falling at a rate that breaks every inflation model ever built. And you will never feel it. Here's why. The Deflation You're Not Allowed to Have Every major central bank on earth operates under a mandate to maintain positive inflation. The Federal Reserve targets 2%. The ECB targets 2%. The Bank of Japan targets 2%. The Bank of England targets 2%. Their entire institutional purpose is to prevent prices from falling. When technology creates deflation — when the cost of producing goods and services falls because AI makes everything cheaper — central banks don't celebrate. They intervene. They expand the money supply. They lower interest rates. They buy bonds. They do whatever is necessary to push prices back up to their 2% target. This is not theoretical. It is explicitly stated policy. And it has profound implications for every asset class on earth. AI pushes the cost of intelligence down by 99.9%. The central bank prints money to push the price level back up. The deflation gets absorbed. Not by consumers — they never see the lower prices. The deflation gets absorbed by asset prices. Stocks, real estate, commodities, Bitcoin — everything with a fixed or scarce supply captures the excess liquidity that central banks inject to fight the deflation that AI is creating. A recent academic paper stated the mechanism plainly: the marginal impact of AI productivity gains is deflationary, but an overall downward impact on prices would require the Fed to fail to undertake an offsetting monetary expansion. In other words, the deflation only reaches consumers if the central bank allows it. And central banks have spent 50 years proving they will never allow it. The Inflation You Will Feel Instead If the deflation doesn't reach you, what does? Asset inflation. The excess money supply that central banks create to offset AI's deflationary pressure flows into scarce assets. The mechanism is straightforward: when you expand the money supply by 7% annually (the long-term M2 growth rate) while AI is simultaneously reducing the cost of production, the gap between money supply growth and consumer price growth has to go somewhere. It goes into assets. This is why the S&P 500 is up 80% since ChatGPT launched in November 2022. It's not because corporate earnings grew 80% in real terms. It's because the monetary system absorbed three years of AI-driven deflation by expanding the money supply, and the excess liquidity flowed into equities. This is why gold is at all-time highs. It's why Bitcoin went from $16,000 to $126,000 in three years. Scarce assets are absorbing the deflation that central banks refuse to let consumers experience. The result is a world that gets cheaper to operate in and more expensive to own in simultaneously. The cost of intelligence falls 99.9%. The cost of scarce assets climbs year after year. Both statements are true. They're not contradictions — they're the same phenomenon viewed from two different positions in the economic hierarchy. The Two Economies AI is creating a bifurcation in the global economy that will define wealth and poverty for a generation. Economy One is the production economy. This is where AI's deflationary force operates. The cost of writing code, generating content, analyzing data, providing customer service, managing logistics, and performing cognitive labor is collapsing. Companies that deploy AI reduce their cost structures dramatically. Margins expand. Earnings grow. The S&P 500 is beating earnings estimates by 4.2x the five-year average because AI is making companies more profitable at a pace the market hasn't seen in decades. Economy Two is the asset economy. This is where the monetary response to deflation operates. As central banks expand money supply to offset the deflation in Economy One, the excess liquidity flows into stocks and scarce assets. Asset prices rise. The cost of owning a stake in the economy climbs faster than wages. If you own assets, you benefit from both economies. Your companies become more profitable (AI deflation) and your assets appreciate (monetary expansion). You compound from both directions. If you earn wages and own no assets, you get neither. Your labor competes with AI that's getting cheaper every quarter. Your wages stagnate or decline in real terms. And the assets you need to buy to build wealth are inflating away from you faster than you can save. The gap between these two positions is the central economic story of the next decade. AI doesn't close this gap. It widens it. Not because AI is bad — because the monetary system's response to AI's deflation systematically transfers wealth from wage earners to asset owners. What This Means for Risk Assets If the framework above is correct, the implications for financial markets are significant and directionally clear. Equities go up structurally. Not because every company deserves a higher valuation, but because the monetary expansion required to offset AI deflation inflates the asset class that absorbs excess liquidity most efficiently. Companies that deploy AI effectively get a double benefit: expanding margins from lower costs and expanding multiples from excess liquidity. The S&P 500's current earnings season — 29.2% aggregate beat, 16.9% net margins at all-time highs — is the first clear evidence of this dynamic at scale. The Mag 7 continue to dominate. These are the only companies growing revenue and earnings at rates that exceed both the M2 growth rate (7%) and the AI-driven margin expansion. They are simultaneously the producers of AI (which gives them the tools) and the deployers of AI (which gives them the margins). Capital concentrates in them not because of speculation but because they are the only equities generating real returns after debasement. AI infrastructure stocks are not a trade — they are the capex layer of a structural shift. $800+ billion in hyperscaler capex this year, projected to exceed $1 trillion in 2027. This spending is building the physical infrastructure of the deflationary engine. The companies supplying it — memory, GPUs, networking, power — are posting record earnings because the buildout is accelerating, not because of a temporary cycle. Bitcoin is the optimal asset in a world of AI deflation plus monetary expansion. Its supply is mathematically fixed at 21 million. It cannot be diluted. It has no issuer who can respond to AI deflation by printing more of it. Every other currency, every other monetary asset, will be expanded in response to AI's deflationary force. Bitcoin is the only monetary instrument that structurally cannot participate in that expansion. As M2 grows to offset AI deflation, Bitcoin's scarcity becomes more pronounced relative to every fiat currency on earth. Gold benefits from the same dynamic but with two structural disadvantages: mining adds roughly 1.5% new supply annually (compared to Bitcoin's fixed schedule approaching 0%), and gold cannot be transferred, divided, or verified at the speed required by a digital economy. Bonds are structurally impaired. Fixed-income instruments that pay 4-5% nominal returns in a world where M2 grows at 7% are guaranteed losers in real terms. The longer the duration, the worse the outcome. Lending money at a fixed rate while the central bank is actively devaluing the unit you'll be repaid in is a bet against the explicit policy of every major central bank on earth. The Central Bank Trap Here's the part nobody is willing to say out loud. Central banks cannot allow AI deflation to reach consumers. Falling prices sound good in isolation — cheaper goods, cheaper services, lower cost of living. But falling prices in a debt-based monetary system are catastrophic. Deflation increases the real value of debt. When prices fall, every dollar of existing debt becomes harder to service. Corporate bonds, sovereign debt — all become more expensive in real terms. The U.S. government has $39.8 trillion in debt. Interest expense is already the largest line item in the federal budget. If AI deflation were allowed to push consumer prices down, the real cost of servicing that debt would increase at exactly the moment when tax revenues (which are tied to nominal economic activity) are falling. Deflation breaks the debt. And the entire modern economy is built on debt. So central banks will fight AI deflation with every tool they have. They will expand the money supply. They will keep rates lower than the deflation warrants. They will ensure that the price level never falls, regardless of how much cheaper AI makes everything to produce. The result is a permanent and widening gap between the cost of production (falling, thanks to AI) and the cost of living (rising, thanks to monetary expansion). The productivity gains accrue to corporations and asset owners. The monetary expansion costs accrue to wage earners and savers. AI makes the pie bigger. Monetary policy decides who gets the extra slices. And the answer, for as long as central banks target positive inflation in a world of AI-driven deflation, is asset owners. The Investment Implication in One Sentence In a world where AI pushes the cost of production toward zero and central banks print money to prevent prices from falling, the only rational strategy is to own scarce assets that cannot be printed. Equities (ownership of AI-deploying companies). Bitcoin (the only monetary asset with a fixed supply). These are the assets that capture the spread between AI deflation and monetary expansion. Everything else — bonds, cash, wages, savings accounts — sits on the wrong side of the equation. The cost of intelligence fell 99.9% in 26 months. The S&P 500 rose 80%. Bitcoin rose 700%. The correlation is not a coincidence. It's the mechanism. AI makes everything cheaper to produce. Central banks make everything more expensive to own. The gap between those two forces is where all the wealth gets created and destroyed over the next decade. This is not financial advice.

    Value Investing Is Slowly Going Broke and Nobody Wants to Admit It

    The real hurdle rate isn't CPI. It's the rate of money supply growth. At 7% annually, most "value" stocks aren't preserving wealth — they're losing it in slow motion. Only growth clears the bar.

    There's a debate in investing that has been running for decades: growth versus value. Graham and Buffett built empires on the value side. The past 15 years have belonged almost entirely to growth. Most people frame this as a style rotation — value will come back, growth is overvalued, reversion to the mean is inevitable. It's not coming back. And the reason has nothing to do with stock picking, sector rotation, or market cycles. It has to do with the unit everything is priced in. The Hurdle Rate Nobody Talks About Every investment has a hurdle rate — the minimum return it needs to generate to justify holding it. Ask most investors what their hurdle rate is and they'll say something like "beat inflation" or "beat the risk-free rate." CPI is running at 4.2%. The 10-year Treasury yields 4.7%. So the hurdle must be somewhere around 4-5%, right? Wrong. CPI is a politically constructed number designed to understate the rate at which your purchasing power is being eroded. It excludes asset prices. It uses substitution adjustments. It uses hedonic quality adjustments. It is, by design, a number that makes inflation look lower than it actually is. The real measure of currency debasement is the rate at which the money supply grows. If the number of dollars in existence increases by 7% per year, then your dollars are being diluted by 7% per year — regardless of what CPI says. U.S. M2 money supply has grown at a compound annual rate of approximately 7% over the long term. That is your actual hurdle rate. Not 2%. Not 4.2%. Seven percent. Any investment that fails to grow its earnings, revenue, or intrinsic value faster than 7% annually is not "preserving capital." It is losing purchasing power in real terms while appearing flat or slightly positive in nominal terms. The statement on your brokerage account says you made 4%. The money supply says you lost 3%. Why Value Stocks Fail the Test The fundamental characteristic of a "value" stock is that it trades at a low multiple to current earnings. Low P/E, low P/B, high dividend yield. The market assigns a low multiple because it expects little to no growth. That's the definition. A company earning $10 per share with 0% growth is earning $10 per share forever. In nominal terms, that looks stable. In real terms — measured against 7% annual money supply growth — that $10 is worth $9.30 next year, $8.65 the year after, and $7.47 within five years. The earnings aren't growing, but the denominator they're measured in is shrinking. After a decade of 0% nominal growth, a value stock's real earnings have declined by roughly 50%. The company didn't do anything wrong. It didn't lose market share. It didn't mismanage capital. It simply failed to grow faster than the rate at which the currency it's denominated in was being debased. This is the structural trap of value investing in a fiat monetary system: you are buying companies whose primary characteristic is the absence of growth, in a currency whose primary characteristic is perpetual expansion. The math doesn't work. It hasn't worked for 15 years. And it won't work as long as M2 continues to grow at 7%. A value stock growing earnings at 3% in a world where M2 grows at 7% is losing 4% per year in real terms. A value stock paying a 4% dividend while growing at 0% is returning negative 3% annually after debasement. The dividend isn't income — it's a partial refund on the purchasing power you're losing by holding the position. Why Growth Stocks Are the Only Ones That Clear the Bar The Mag 7 — Apple, Microsoft, Google, Amazon, Meta, Nvidia, Tesla — have dominated the S&P 500 for a reason that has nothing to do with momentum trading or retail speculation. Every single one of them grows revenue and earnings at rates that exceed 7% annually. Most grow at 15-40%+. Nvidia grew revenue 345% in a single year. Microsoft grew Azure to $100 billion. Meta tripled its AI revenue. These companies aren't just beating inflation. They're beating the debasement. When a company grows earnings at 25% per year in a world where M2 grows at 7%, the real earnings growth is roughly 18%. The company is genuinely creating value that outpaces the dilution of the currency. That's real wealth creation. That's what shows up not just on the brokerage statement but in actual purchasing power. This is why the Mag 7 have absorbed an ever-increasing share of total market capitalization. It's not a bubble. It's the market correctly identifying the only companies whose growth rates exceed the rate of monetary debasement. Capital concentrates in the assets that are actually gaining ground in real terms. The S&P 500's CAGR of 14% over the past five years looks impressive until you subtract the 7% M2 growth rate. Real return: roughly 7%. Still good. But the Mag 7 components within the index are doing 20-40%. The other 493 stocks are collectively barely keeping pace with the money supply. The index return is being carried almost entirely by growth. CPI Is Not Your Benchmark This is the point that most investors — including sophisticated ones — refuse to accept. CPI measures a basket of consumer goods and services. It tells you how much more expensive groceries and gasoline are this year versus last year. It is useful for that narrow purpose. It is useless as a measure of how fast your wealth is being diluted. Asset prices are not in CPI. Your house is not in CPI. The S&P 500 is not in CPI. Bitcoin is not in CPI. The things that actually store wealth are excluded from the metric that supposedly measures its erosion. When M2 grows at 7% and CPI reads 4.2%, the gap isn't a rounding error — it's the portion of debasement flowing into asset prices instead of consumer prices. That gap is why houses cost 5x what they did 20 years ago while CPI cumulated roughly 60% over the same period. The money supply more than tripled. Consumer prices didn't. The difference went into assets. If you benchmark your portfolio against CPI, you're measuring yourself against a number that structurally understates the rate at which your wealth is being eroded. You can "beat inflation" by CPI standards while still losing ground in real purchasing power terms. That's not a theoretical concern — it's the lived experience of every saver and value investor for the past two decades. The Mag 7 Through the Debasement Lens Look at each Mag 7 company through this framework: Apple: services revenue growing 15%+ annually with a $3.8 trillion market cap that continues to expand. The company grows faster than M2 and returns capital through buybacks that reduce share count — double-compounding against debasement. Microsoft: Azure growing at 30%+ and crossing $100 billion in annual revenue. The AI capex cycle is adding a new growth engine on top of an already dominant cloud and enterprise software business. Nvidia: revenue growth measured in hundreds of percent. The company is so far ahead of the debasement rate that it exists in a different category entirely. Meta: pivoted from a metaverse spending narrative to an AI revenue machine. Growth reaccelerated. The market repriced it from $88 to $740 once the growth rate was confirmed. Amazon: AWS and advertising revenue growing at 20%+. Retail is the loss-leader for two high-margin businesses that compound well above 7%. Google: search and cloud revenue growing at 15-20%. YouTube alone is a $50 billion business. AI capex is building the next growth layer. Tesla: automotive revenue growing at 20%+ with energy and AI optionality providing additional growth vectors. Every one of these companies is growing earnings at multiples of the M2 growth rate. That's not coincidence. That's the market systematically concentrating capital in the only equities that generate real returns after debasement. What This Means for Portfolio Construction If M2 is your real hurdle rate, the implications for how you build a portfolio are stark: Bonds returning 4-5% are guaranteed losers in real terms. You're locking in a return below the debasement rate for 10-30 years. The coupon doesn't compensate for the purchasing power you're losing. Value stocks growing at 0-3% are slow-motion wealth destruction dressed up as "conservative investing." The dividend yield looks like income. It's actually a partial offset on a losing position. Cash is the worst-performing asset class in a 7% M2 growth environment. Every dollar you hold loses 7% of its purchasing power annually. In five years, your $100 buys $70 worth of goods — but that's CPI. In asset terms, it buys $50 or less. Growth stocks growing at 15%+ are the only equity class that reliably outpaces M2. The concentration of capital in the Mag 7 isn't irrational exuberance — it's rational capital allocation in a debasing monetary environment. Bitcoin, with a 50% CAGR over the past five years and a fixed supply of 21 million, is the ultimate anti-debasement asset. It doesn't just outpace M2 — it exists entirely outside the system that produces M2. The Value Trap Is Structural Value investing worked in a world where the money supply grew slowly and predictably. When M2 grew at 2-3% annually, a company growing earnings at 5% with a 3% dividend was genuinely creating wealth. The hurdle was low. Many companies could clear it. That world ended decades ago. M2 growth accelerated. Deficits expanded. Central banks printed. The hurdle rate climbed from 3% to 7% — and the vast majority of "value" companies couldn't keep up. They didn't change. The monetary system changed around them. The value investing community has spent 15 years waiting for mean reversion that isn't coming. They're waiting for the market to "realize" that low P/E stocks are cheap. The market already realizes it. Low P/E stocks are cheap because the market correctly prices in that their growth rate doesn't clear the debasement hurdle. The low multiple isn't an opportunity — it's a verdict. Growth will continue to outperform value for as long as central banks target positive inflation and money supplies continue to expand. That's not a cycle. That's a regime. And the regime isn't changing. The Bottom Line The real hurdle rate for any investment is the rate of money supply growth — approximately 7% annually for the U.S. dollar. CPI understates the true rate of debasement by excluding asset prices. Any company growing slower than 7% is losing ground in real terms, regardless of what the nominal numbers say. This is why value stocks have underperformed for 15 years and will continue to underperform. Their defining characteristic — low growth — is the one thing that guarantees failure in a monetary system built on perpetual currency expansion. The Mag 7 dominate because they are the only large-cap equities consistently growing at multiples of the debasement rate. Capital isn't concentrating in growth stocks because of speculation. It's concentrating there because growth is the only strategy that works when the money is broken. The debate between growth and value isn't a style question. It's a monetary question. And the monetary system has already answered it. This is not financial advice.

    MiniMax Is Up 20% in a Week and 82% Off Its Highs. One of Those Numbers Is About to Change.

    China's most aggressive AI lab just launched a model that 100+ partners integrated within 24 hours. Goldman has a target 3.7x higher than the current price. Stock Connect inclusion could hit this month. And the stock is still 82% below its January high.

    MiniMax Group is doing something unusual in this market: going up while everything around it figures out whether the AI trade is alive or dead. The stock is up 7.7% in the last 24 hours and roughly 20% over the past week, trading around HK$230-260. For context, this is a company that hit HK$1,330 in January and traded as low as HK$186 in July. The 82% drawdown from peak to trough erased almost the entire post-IPO premium. What's left is a company growing revenue at triple-digit rates, launching competitive foundation models, and trading at a fraction of what every major investment bank says it's worth. Here's why the snapback might just be getting started. What MiniMax Actually Is MiniMax is one of China's six "AI Tigers" — the cohort of foundation model companies that emerged from China's AI lab boom alongside DeepSeek, Zhipu AI, Moonshot, Baichuan, and 01.AI. Founded in 2022 and headquartered in Shanghai, it listed on the Hong Kong Stock Exchange in January 2026. The company operates across four product lines. Hailuo AI is a video generation platform competing directly with Runway, Pika, and Sora. Talkie (and its Chinese counterpart Xingye) is an AI character companion app with over 11 million monthly active users — one of the most popular consumer AI products on earth. MiniMax Audio handles voice synthesis and music generation. And the core platform is an enterprise API serving foundation models across language, video, speech, and multimodal tasks. Annual recurring revenue hit $400 million in April 2026. Management has publicly stated confidence in reaching $1 billion in ARR by the end of the year. For a company that didn't exist three years ago, the revenue trajectory is moving at a pace that matches or exceeds early-stage OpenAI. The H3 Launch Changed the Momentum The immediate catalyst for the rally is H3 — MiniMax's next-generation multimodal generative AI model, launched last Friday. Within 24 hours of release, more than 100 domestic and international partners completed Day 0 integration. That's not a press release metric — it's a signal that the developer ecosystem is treating MiniMax as essential infrastructure. Day 0 adoption at that scale means partners had been preparing for the launch, which means the model's capabilities were already validated in private before the public release. The market interpreted this correctly. MiniMax surged 15.1% on the day of the H3 launch and another 12.8% in the session that followed. The move wasn't speculative retail enthusiasm — it was institutional repricing of commercial viability after a concrete product milestone. H3 is multimodal — meaning it handles text, image, video, and audio in a single unified model. The competitive landscape for multimodal foundation models is thin. OpenAI has GPT-5, Google has Gemini, and Anthropic has Claude. In China, DeepSeek and MiniMax are the frontrunners. H3 positions MiniMax as a serious competitor in both the Chinese and international markets simultaneously. The Analyst Gap Is Absurd The disconnect between where MiniMax trades and where analysts say it should trade is one of the widest in global equity markets right now. Goldman Sachs: HK$860 target. That's 3.7x the current price. Citi: HK$533 target. Approximately 130% upside. Bank of America: Buy rating with detailed analysis of MiniMax's computing cost advantages and overseas cloud partnerships. The average 12-month target across 14 Buy-rated analysts is HK$766.80 — implying 233% upside from current levels. In early July, Goldman, BofA, and Citi issued simultaneous Buy ratings — a rare coordinated signal that typically precedes significant institutional positioning. Goldman's report specifically highlighted the Chinese AI pricing environment shifting toward rationalization, with competitor DeepSeek introducing differentiated peak-hour pricing. This signals the end of the aggressive price war that compressed margins across China's AI industry since late April, which is directly positive for MiniMax's API revenue quality. Stock Connect Is the Structural Catalyst Goldman Sachs expects MiniMax may qualify for inclusion in the Hong Kong Stock Connect program in August 2026. Stock Connect is the mechanism that allows mainland Chinese investors to purchase Hong Kong-listed stocks through the southbound trading link. Inclusion would open MiniMax to hundreds of millions of potential Chinese retail and institutional investors who currently cannot access it. Estimates suggest Stock Connect eligibility could bring tens of billions of Hong Kong dollars in fresh inflows into the stock. That's not a one-time event — it's a structural expansion of the buyer base that permanently increases liquidity and creates ongoing demand from mainland funds. For a stock that crashed 82% partly due to thin liquidity and concentrated selling, Stock Connect inclusion would fundamentally change the supply-demand dynamics. More buyers, deeper order books, and a structural bid from mainland capital that didn't exist before. The timing — potentially this month — aligns with the H3 launch momentum and the upcoming M3 Pro debut. If inclusion happens during a period of positive product catalysts and improving analyst sentiment, the repricing could be violent. The M3 Pro Model Is Next The H3 launch is the appetizer. M3 Pro is the main course. Expected between September and October 2026, M3 Pro will be MiniMax's flagship model with approximately 2.7 trillion parameters — approaching the 3 trillion parameter frontier. The technical approach includes sparse attention mechanisms, activation parameter optimization, and KV Cache optimization designed to deliver frontier-level capabilities at competitive inference costs. MiniMax's current M3 model offers blended pricing of $0.22 per million tokens. For context, recall Chamath's pricing chart: Anthropic charges $56 per million tokens, OpenAI charges $26, and Chinese models are at $0.50 or below. MiniMax is competing at the bottom of the cost curve with performance approaching the top. If M3 Pro delivers on its promise of frontier capabilities at Chinese model pricing, it becomes the most cost-efficient high-performance foundation model available globally. Enterprise customers who currently pay 100x more for comparable capabilities from Western labs would have a compelling reason to evaluate MiniMax — particularly for inference-heavy workloads where cost per token determines unit economics. 82% Off Highs With This Catalyst Stack Step back and look at the full picture. MiniMax is an AI foundation model company with: $400 million ARR growing toward $1 billion by year-end. A multimodal model (H3) that 100+ partners integrated on day one. A 2.7 trillion parameter model (M3 Pro) launching in two months. Three simultaneous Buy ratings from Goldman, BofA, and Citi. An average analyst target implying 233% upside. Potential Stock Connect inclusion this month. An industry pricing environment shifting from destructive price war toward rationalization. Consumer products with 11 million+ monthly active users. A $2 billion funding round recently closed. And it trades 82% below its January high. The AI selloff hit everything. MiniMax, like every AI stock on earth, got caught in the broad rotation out of the sector during the KOSPI crash, the Leopold liquidation, and the Nasdaq correction. But the company's product trajectory didn't slow down while the stock was falling. H3 launched. Partners integrated. ARR kept climbing. The business accelerated while the stock collapsed. The China AI Divergence For Stockgecko readers tracking both sides of the AI trade, MiniMax represents the clearest example of the divergence we flagged last week: Chinese AI names ripping while American AI names were getting destroyed. While SK Hynix was down 55%, Nvidia was correcting, and Leopold was getting margin called out of U.S. semiconductor positions, Chinese AI companies were posting gains. CXMT debuted with a 500% pop. MiniMax rallied 20% in a week. The capital isn't leaving AI — it's rotating into the part of AI that the selloff made cheapest. The valuation gap is stark. U.S. AI infrastructure companies trade at elevated multiples even after the correction. Chinese AI companies trade at fractions of their analyst targets after an 80%+ drawdown. The risk is higher — regulatory uncertainty, geopolitical exposure, and China-specific market dynamics are all real. But the asymmetry at current prices is hard to ignore. The Risks The 82% drawdown happened for reasons that haven't fully resolved. Liquidity is thin. Pre-IPO shareholder lockup dynamics create overhang risk. The stock has a 52-week range of HK$186 to HK$1,330 — that kind of volatility isn't for everyone. China's AI pricing environment remains brutal. Even with signs of rationalization, margins are compressed across the industry. MiniMax needs to prove it can grow ARR to $1 billion while maintaining or improving API gross margins — and that's not guaranteed in a market where DeepSeek regularly undercuts competitors. Geopolitical risk is constant. U.S.-China tensions, export controls on computing hardware, and potential regulatory actions against Chinese AI companies operating internationally are all headwinds that can't be modeled or predicted. And Stock Connect inclusion isn't confirmed. Goldman expects eligibility in August, but final inclusion depends on exchange decisions that MiniMax doesn't control. If it's delayed, the near-term catalyst weakens. The Bottom Line MiniMax is 82% off its highs with triple-digit revenue growth, a multimodal model that 100+ partners adopted on launch day, a 2.7 trillion parameter model launching in two months, three simultaneous Buy ratings from major banks with an average target implying 233% upside, and potential Stock Connect inclusion this month. Either 14 analysts are simultaneously wrong about a company growing toward $1 billion in ARR, or the stock is one of the most mispriced AI names on earth. Track MiniMax and every HIP-3 asset on Stockgecko. Prices and data referenced as of August 5, 2026. This is not financial advice.

    Data sourced live from Hyperliquid. Refreshes every 15 seconds.