Wall Street can really be quite double-standard sometimes. Some companies get their stock price cut as soon as they say they want to spend more on AI. Amazon said it plans to spend $220 billion this year, yet $AMZN actually rose 15.3% in one day, with its market cap nearly hitting $3 trillion. Normally, if free cash flow goes from an inflow of $18.2 billion to an outflow of $7.6 billion, people would run away first. The difference of $25.8 billion in and out is more money burned than many companies make in a year. But the market not only didn’t run, it applauded collectively. The reason lies in AWS. In Q2, AWS revenue was $42.2 billion, up 36.7% year-over-year, the fastest growth in 18 quarters. Operating profit was $16.6 billion, with a profit margin close to 39%. Amazon’s total operating profit for the quarter was only $27.5 billion, with 60% contributed by AWS. The money Amazon is now pouring into AI data centers has already started to come back from customer bills. AWS’s AI business and self-developed chip business each have annualized revenues exceeding $25 billion. Jassy even said that after spending $220 billion this year, capacity is still insufficient, and demand for 2028 is already very strong. The machines aren’t built enough yet, but customers are already lining up at the door. There’s also a number that’s easy to misinterpret. Amazon’s Q2 net profit was $62.6 billion, which includes $53.4 billion in pre-tax non-operating income, mostly from the Anthropic investment. If you treat the entire $62.6 billion as Amazon’s regular business earnings, you’d get the wrong picture. Looking at how the company itself performed, the $27.5 billion operating profit is more reliable. So this rise is quite reasonable. Wall Street has never feared companies spending money; it fears that after the money is spent, nothing happens. Amazon is currently at the stage where the money has just been spent, and AWS’s revenue and profits are already climbing. The burn rate is scary, but the collection rate hasn’t faltered. Of course, $220 billion isn’t a small amount. If AWS’s growth slows down too much in the future, or if profit margins are squeezed by depreciation and chip costs, what everyone praises today as foresight could be criticized tomorrow as overspending.

I'm curious about one question: Apple's earnings report is clearly quite good, so how could $AAPL still drop 7% in one day? $AAPL closed at $308.91 on Friday, down 7.1%. Revenue was $109.4 billion, up 16% year-over-year; EPS was $2.02, up 29%; iPhone revenue was $54.3 billion, up 22%. This can't be considered bad, you could even say it sold extremely well. I then looked into the inventory and next quarter guidance, and when I saw the inventory line, I kind of understood. Apple's quarter-end inventory was $11.1 billion, nearly double last September's $5.7 billion. Next quarter revenue growth guidance is only 9% to 11%, and gross margin is expected to drop from this quarter's 50.1% down to 47% to 48%. iPhone sales don't seem to be a problem at the moment. Mac revenue was $10.4 billion, up 29%; services revenue was $30.7 billion, also up 12%. What people are worried about now is how much profit Apple can keep from each unit sold. AI giants have been aggressively building data centers these past two years, driving memory prices higher and higher. Apple hasn't spent hundreds of billions in a single quarter on AI data centers like Microsoft, Google, or Meta, but still ended up paying an AI tax from afar. Cook himself said memory price increases have become extreme. Mac and iPad prices have already gone up, but iPhone prices haven't changed yet. This makes things difficult. If they don't raise prices, Apple has to absorb the cost; if they do, fewer people might upgrade. Whether Siri and on-device AI can actually motivate people to buy new devices hasn't been fully proven yet. But the memory bill has already arrived. Doubling inventory isn't necessarily all bad. Earlier demand was better than Apple expected, and supply has tightened, so stocking more is normal. As long as this inventory sells smoothly and gross margin holds, this drop might really be an overreaction. The real fear is if inventory keeps piling up and gross margin keeps falling. Then it would become a situation where iPhones sell like crazy, but Apple’s profits aren’t as comfortable as before.

Coinbase $COIN's Q2 earnings report came out, and after-hours it once dropped nearly 7%. After reading it, it's actually not undeserved. Over the past few years, Coinbase has developed stablecoins, custody, subscriptions, and prediction markets, always trying to shed the label of "making money solely from users trading crypto and paying fees." But as crypto trading cooled down, revenue immediately dropped. Q2 total revenue was $1.22 billion, while the market originally expected about $1.29 billion. Trading revenue fell from $756 million last quarter to $599 million. That's about $157 million less earned in three months. More awkwardly, Coinbase's market share actually rose from 9.1% to 10.3%. Simply put, a higher proportion of people are trading on Coinbase, but the entire crypto market is less active. The overall pie shrank, so even with a bigger share, the fees collected are still less. Subscription and service revenue was $555 million, accounting for 48% of net revenue, which seems to already support half the business. But this revenue also dropped 5% quarter-over-quarter. The higher share partly results from trading revenue falling faster. $USDC is similar. The average $USDC held in Coinbase products reached $20 billion, a new high, but stablecoin revenue dropped from $305 million to $292 million. More money is held, but with interest rates falling, Coinbase still earns less. The prediction market, however, grew rapidly, with contract numbers and revenue more than doubling from last quarter, annualized revenue exceeding $100 million. But $100 million annualized spread over a quarter is about $25 million. Trading revenue dropped $157 million in one quarter, so this new income can't cover it yet. So Coinbase's current problem is clear. The new businesses are launched and indeed used, but the earnings are not enough yet. Adjusted EBITDA is still $208 million, positive for 14 consecutive quarters. The company booked a $360 million loss, including about $210 million in crypto investment losses and over $52 million in restructuring costs. The core business hasn't suddenly deteriorated. But $COIN has already risen a lot before, and people bought into the story of "making a lot of money without relying on trading fees in the future." Now the story exists, but the money hasn't fully caught up. Coinbase wants to change its business model, and the new businesses are all in place. But as long as $BTC has no momentum, the earnings report will still look bad immediately.

After Microsoft $MSFT released its Q4 earnings report, as of the after-hours quote recorded on Microsoft's investor relations page in the Eastern US, the stock price was around $425, about 8.8% higher than the regular close. But the point where I paused while reviewing the data was actually another set of numbers appearing side by side: this quarter's revenue of $90 billion, and capital expenditures plus finance leases of $41 billion. One reflects the company's speed of making money, the other the speed at which the company continues to invest heavily in AI and data centers. In recent quarters, Microsoft's biggest issue wasn't "whether it can do AI." The real market bottleneck was whether the money spent could turn into revenue, rather than just staying stuck on GPUs, data centers, and the OpenAI relationship. Yesterday's earnings report at least provided a firmer answer than previous quarters. Azure and other cloud services revenue grew 43% year-over-year, exceeding the market's original expectation of around 40%. Intelligent Cloud revenue was $39.3 billion, up 32% year-over-year. Microsoft's broad cloud revenue was $59.3 billion, up 27% year-over-year. If you only say "cloud business is strong," these numbers would be understated. At first, I thought the after-hours surge was mainly due to Azure's impressive growth, but when I looked at the cash flow statement, I realized the market was actually looking at both sides simultaneously. Free cash flow was still $19.6 billion but down 23% year-over-year; in the same quarter, capital expenditures and finance leases jumped 69% year-over-year. This is not easy growth; it is very expensive growth. Microsoft can temporarily get investors to accept this spending because of two things that have already translated into revenue. Azure's full fiscal year revenue exceeded $100 billion for the first time, and Microsoft 365 Copilot paid seats surpassed 30 million, up from just over 20 million in April. In other words, AI is finally not just a demand talked about on conference calls but is starting to appear in cloud bills and software seats. AP quoted Zacks investment strategist Bryan Hayes saying, roughly, that the market is willing to believe this time that Microsoft’s spending has bought something real. I think this statement closely matches the after-hours reaction. Previously, people feared AI Capex was bottomless; now seeing revenue, orders, and seats all rising together, doubts have been pushed back a notch. Also, don't overestimate net profit growth. Microsoft's Q4 net profit was $35.8 billion, GAAP EPS $4.81, but this included a $3.2 billion gain from Anthropic investment and non-GAAP adjustments from OpenAI investment. The company’s operations are indeed strong, but the investment scope also makes the quarterly numbers more complex. The management’s guidance for the next quarter also explains why $MSFT surged after hours. Microsoft expects next quarter revenue to be roughly $89.85 billion to $90.95 billion, with Azure growing 45% at constant currency, exceeding StreetAccount’s expectations. The capital expenditure plan has not been increased further, which also relieved the market. What really changed my view in this earnings report is that Microsoft broke down the phrase "AI is expensive" into a calculable account. It is indeed expensive, with $41 billion in one quarter; but if Azure continues to grow over 40%, and Copilot seats keep increasing, investors are willing to give it time. Conversely, if Azure’s growth rate falls back, or Copilot seat growth slows, and free cash flow continues to be pressured by Capex, the market’s patience will shorten. What this earnings report has bought $MSFT is not a free pass but time to prove next quarter that this AI investment can turn into cash.

Minnesota did not ban prediction markets, and this is really related to Robinhood $HOOD Minnesota originally planned to ban prediction markets starting August 1, but before the law took effect, it was blocked by a federal judge. For Robinhood $HOOD, the fastest-growing product line in company revenue is prediction markets. Robinhood did not participate in this lawsuit. The CFTC sued Minnesota, with Kalshi and Polymarket later joining. On July 27, the judge issued a preliminary injunction, believing they had a strong chance of winning. During the ongoing litigation, the state government is temporarily barred from enforcing this ban. Why is Robinhood involved? Its event contracts are provided by Robinhood Derivatives, and trading takes place on a CFTC-regulated partner exchange. Minnesota wants to use state gambling laws to ban this type of product, but Robinhood follows the federal regulatory path. The judge temporarily accepted the CFTC's argument, naturally reducing one layer of trouble for Robinhood. This business is already sizable. In Robinhood's May data release, users traded a total of 3.9 billion event contracts, 22% more than in April; the daily average was 126 million contracts, up 18%. The 3.9 billion figure refers to the number of contracts, not $3.9 billion in transaction volume. Each contract is priced between $0.01 and $0.99, and settles at $1 or $0 after the event outcome. The contract count looks large but is not comparable to stock trading volume. Robinhood has also started charging fees on these trades. Starting June 1, the platform charges commissions based on contract price, up to $0.01 per contract, with exchanges possibly charging an additional $0.01. The 3.9 billion contracts cannot be directly multiplied by one cent to calculate revenue, as actual commissions vary with price and account tier, but generally, higher volume means more fees collected. Last November, Robinhood said that one year after launching prediction markets, over 1 million users had traded 9 billion contracts, calling it the fastest-growing product line by revenue. The company later formed a joint venture with Susquehanna, acquiring a CFTC-regulated exchange and clearinghouse, preparing to move more trading onto its own infrastructure. Wall Street is already raising expectations for this business. Needham's John Todaro recently raised Robinhood's earnings forecasts and target price, explicitly citing acceleration in stocks, options, and event contracts. KeyBanc also raised its target price but cautioned that after a period of stock price gains, Robinhood is not its favorite pre-earnings trade. Both sides have practical concerns. The business is indeed growing fast, and the stock price already reflects much of that growth. Robinhood will release Q2 earnings after market close on July 29, with market expectations around $1.29 billion in revenue and $0.43 earnings per share. If prediction markets only look good in contract volume but revenue contribution is unclear, investors may not continue to price it at high growth. This ruling does not resolve the state's issues once and for all. It is only a temporary injunction; Minnesota will continue the lawsuit. Robinhood's own help page states that event contracts are not available in every state; Maryland currently does not allow them, and Nevada users cannot open new sports contracts. This lawsuit is a positive for $HOOD, though not significant enough to solely determine the stock price. It preserves Robinhood's current business path, saving some regulatory friction. How much fee revenue and how many new users come from the 3.9 billion contracts will be clearer in tomorrow morning's earnings report numbers.

The South Korean stock market fell 10.84% in one day, and the AI chip calculations are being reconsidered Today, the South Korean stock market really stunned me. The KOSPI closed with a sharp drop of 10.84%, with an intraday maximum decline of 11.29%, triggering a circuit breaker. Samsung Electronics fell 13.39%, and SK Hynix dropped 14.65%. Japan's Nikkei 225 index also fell nearly 4%, and Taiwan's weighted index dropped 4.7%. The stocks sold off the hardest today were basically the hottest AI and semiconductor stocks from the past year. At first, I thought the main concern today was the market worrying about US tech companies burning too much cash. After reading all the news, there are several more direct triggers for this South Korean plunge. ChangXin Memory Technologies surged on its IPO day, causing the market to worry again about competition from Chinese memory chip companies. A report about China's domestic DUV chip equipment starting mass production further fueled this concern. Additionally, Samsung and SK Hynix had risen too much earlier, and South Korea has many leveraged ETFs tracking these two companies' stocks. Once the market starts to fall, leveraged products amplify the volatility. So this round of sharp decline today is hard to explain with just one reason. But when AI capital expenditures will break even has indeed become an unavoidable question for the market. In the past two years, whenever Google, Microsoft, or Meta announced increased data center and chip investments, the market usually got very excited, thinking AI demand was about to explode again. Now, when people see such news, their first reaction is: why keep spending? How much more will be burned? When will it break even? Alphabet $GOOGL just raised its capital expenditure forecast for this year last week to $195 billion to $205 billion. Capital expenditure in Q2 reached $44.9 billion, and free cash flow turned negative $5.9 billion. In the same quarter, Google Cloud revenue grew 82%, indicating that AI and cloud business demand is indeed still there. It's just that the money is being spent faster than cash is earned back. The company's business hasn't suddenly worsened; the market just isn't as willing to wait as before. Previously, as long as AI demand was proven, investors were willing to give high valuations. Now, it is necessary to continue proving that these demands can ultimately turn into profits and cash. Samsung and SK Hynix falling so much today doesn't mean the two companies deteriorated by double digits in one day. They are at the core of the AI storage industry chain, had risen a lot before, and positions were crowded. Now, combined with competition from Chinese manufacturers, overvaluation, and leveraged funds reducing positions, their stock prices naturally take the hit first. Today's market also incidentally shows that when tech stocks plunge, crypto assets find it hard to remain completely unaffected. In the Asian morning session, $BTC once dropped 2.3% to around $63,414, and $ETH fell 3.6%. When funds really want to reduce risk, they usually sell the most liquid assets first. At this time, chip stocks and mainstream coins can easily be put into the same sell basket. Whether this wave can stabilize still depends on the business itself. Whether Samsung and SK Hynix's orders have decreased, whether storage prices and profit margins will loosen; whether the money spent by Google, Microsoft, and others can gradually be earned back from cloud business and AI products. If orders don't drop and cash flow slowly catches up, then today looks more like a concentrated bubble squeeze after too much prior rise. If Chinese manufacturers continue to seize the market, AI capital expenditures keep burning more and more, and the break-even time keeps getting pushed back, chip stock valuations will have to be recalculated downward, and the crypto circle can't expect to completely avoid it. It's still too early to say AI demand is failing. What is certain is that the market is no longer as easy to please as before. Everyone paid a high price for AI earlier, and now they are starting to urge these companies to show their report cards.

BlackRock has also started pushing the CLARITY Act; Wall Street really can't wait anymore Recently, BlackRock publicly supported the CLARITY Act. My first reaction was: Wall Street really can't wait. The crypto community has been calling for "regulatory clarity" for many years. Project teams and exchanges have been shouting about it every day, which I think is normal—no one wants to suddenly receive a subpoena from the SEC. But now even BlackRock is personally stepping in to push this, which makes it a bit different. Samara Cohen, BlackRock's Global Head of Market Development, recently publicly supported the CLARITY Act. What she basically means is: set the rules quickly, protect investors, but don’t drive all new market opportunities outside the U.S. What exactly is the CLARITY Act about? Simply put, it aims to clearly define which regulator oversees which crypto asset. Currently, many cryptocurrencies have been listed in the U.S. for years, and even project teams themselves can't clearly say whether they are securities or commodities. The SEC thinks it should regulate them, the CFTC has its own jurisdiction, and in the end, people often have to wait for lawsuits and court rulings to find out. CLARITY wants to establish a basic division of responsibilities: Projects that raise funds by selling tokens will mainly be regulated and disclosed by the SEC; mature blockchain digital commodity spot markets will be more under the CFTC. Exchanges and brokers must also register. How customer funds are held, whether they can be commingled with the platform’s own funds, and who is responsible if something goes wrong—all must be clearly defined in advance. So don’t misunderstand this as the U.S. preparing to open up crypto trading. There may be more rules in the future, but at least we will finally know who to approach and what standards to follow. Why is BlackRock in such a hurry? It’s actually easy to understand. Their $BTC and $ETH products are already being sold, and they have also launched tokenized funds like BUIDL. The products are in the market, the money has come in, but the underlying rules are still changing back and forth. For a big institution like BlackRock, strict regulation can be considered a cost. What’s really painful is when the rules change every few months because every product they launch has to consider whether it might suddenly be shut down. This bill has not yet been officially enacted. Last year, it passed the House with a vote of 294 to 134. In May this year, the Senate Banking Committee moved it forward, but the full Senate vote and other procedures have not been completed. Senate Majority Leader Thune recently said it might not make it before the August recess. BlackRock stepping up now seems to be urging Congress: stop debating and set the rules quickly. As for whether this news will immediately pump the market, I think that’s wishful thinking. What it really affects is which coins U.S. exchanges dare to list in the future, how projects raise funds through token issuance, and whether Wall Street dares to move more assets onto the blockchain. Some small altcoins might suffer even more because disclosure and compliance cost money, and many projects may not be able to afford it. I think the most interesting part of this is: Wall Street used to be just studying whether it could enter the crypto space. Now they are already in and are complaining that the "renovation" is too slow. BlackRock pushing for rules shows that they see even more business opportunities ahead.

Tesla $TSLA sold 480,000 cars this quarter, generating $28.2 billion in revenue, a record high. However, after the earnings report was released, the stock price dropped 14.5% in one day, wiping out about $203 billion in market value. The reason can be seen by looking at the income statement. Revenue grew 26% year-over-year, but operating profit was only $398 million, down 57% year-over-year. The overall operating margin fell from 4.1% in the same period last year to 1.4%. Free cash flow also turned from positive $1.44 billion last quarter to negative $1.09 billion. Quarterly capital expenditures doubled to $5.8 billion, up 142% year-over-year, already exceeding the company's cash generation speed. This money was mainly invested in Robotaxi, Optimus, AI computing power, chip factories, and new production lines. Tesla's CFO also said that capital expenditures will continue to grow over the next two to three years, with this year expected to exceed $25 billion. Yesterday I wrote about Google $GOOGL, whose $44.9 billion quarterly capital expenditure is already very exaggerated, but Google at least still has two cash machines: search ads and cloud business. Tesla's current situation is a bit more precarious. The car business is responsible for making money, Robotaxi and Optimus are for the future, but the car profits are getting thinner, and the latter businesses have not yet brought in enough revenue. Wall Street is debating this issue now. Morgan Stanley believes investors need to see that this money can truly strengthen Tesla's advantage in "physical AI." After the earnings report, they lowered Tesla's 2026 EBITDA forecast by 7% and expect free cash flow to possibly outflow about $14 billion in 2027. Morningstar is relatively optimistic, believing that short-term cash flow pressure is already reflected in the valuation, and Robotaxi, FSD, and autonomous trucks still have huge commercial potential in the future. Both sides have valid points, but it ultimately comes down to the balance sheet. Tesla is currently spending real money that exists today to buy the future of Robotaxi, Optimus, and AI. Once this future is realized, $TSLA could still be one of the most important companies in the AI era. If the realization speed can't keep up with the spending speed, the market won't keep waiting for Musk forever.

Many times, the crypto industry is a profit-driven sector. Many project teams usually won't invest too many resources to help users reduce risks like theft, mis-signing, or mis-authorization if they don't see direct benefits. In the earlier years, before the OKX wallet existed, operating was really nerve-wracking. If you clicked on a wrong link, connected to a fake website, or signed an unsafe authorization, most of the time you could only blame yourself. Because once an on-chain transaction happens, there's no customer service to help you reverse it. Security is always the user's own responsibility. But after the OKX wallet appeared, the user experience improved countless times. For example, it directly blocks access to risky websites. It warns you when opening high-risk apps. When you're about to sign a dangerous authorization, it highlights the risks. It also flags suspicious addresses in advance. This is basically a nanny-style service. And I think its biggest contribution to the industry is lowering the barrier to entry for web3. I've always felt that for Web3 to truly reach more ordinary users, it can't just rely on the phrase "you manage your own assets." Ordinary people don't understand contracts or hexadecimal data, and they can't possibly do a security audit before every signature. I remember when I first entered the space in 2018-2019, I struggled a lot to figure out how to use MetaMask. Honestly, if the OKX wallet had existed back then, my $ETH wouldn't have been stolen because I clicked on an unsafe contract, haha.