You know that sinking feeling when an annual app subscription auto-renews on your credit card? For two years, Wall Street happily ran on the free trial of artificial intelligence. This week, the invoice arrived.
Two separate bills landed at the exact same time. The first arrived in hardware. Forecasters at TrendForce and GF Securities cut their memory projections, warning of an outright NAND flash glut by 2027. Spot prices dropped over 20%, dragging storage and memory makers down double digits in a single session. Memory is the basic flour in every data center bakery. When flour prices break, you find out fast who was baking artisan bread and who was simply marking up raw grain.
The second bill came from the bond market. Heading into the December Federal Reserve meeting, traders priced in roughly 97% odds of an interest rate hike. Target rates already sit near 4.00%, and central bank governors openly discussed half-point moves to knock down lingering inflation. That matters because higher rates punish companies that promise profits in the distant future.
The tape broke. The S&P 500 sank roughly 8.6% into its first correction in twenty months. Even Nvidia's blowout $115 billion revenue haul could not keep its stock from tumbling 23% across the week. Meanwhile, Anthropic completed the largest public offering in market history, only to give back most of its first-day gain by Friday afternoon. Three of our four names ended the week sitting within inches of their 52-week lows. Our job is separating the temporary bills from the broken franchises.
First, we settle last week's scorecard. Three of those calls were Accumulate; on AMD we said wait for a better entry, which turned out to be the one piece of judgment worth claiming. Equal-weighted, the four names dropped roughly 12.2% against the broader market's 8.6% decline. We held AI-adjacent beta into the exact week that beta unraveled, and we own that miss plainly.
Constellation Energy dropped roughly 15.3%. The long-term nuclear power thesis never wavered, and analysts even raised their targets on data center demand, but the broader sell-off dragged the equity down anyway. Being fundamentally right does not protect you from a falling tape.
RTX proved to be our best decision, falling roughly 4.2%. Its defense contracts provided real ballast while pure technology stocks cratered.
Netflix broke, falling roughly 8.7% to its lowest level since the summer of 2024. Even front-runner status for World Cup media rights could not stop analysts from downgrading its valuation multiple.
AMD suffered the worst damage, plunging roughly 20.5%. A short-seller report alleged unverified server firmware diversions into blacklisted Chinese channels, and the market sold first before asking questions. The tape always delivers the verdict before the facts arrive.
1. Broadcom (AVGO) — The Bespoke Tailor
Nvidia sells off-the-rack luxury suits to anyone with cash. Broadcom is the bespoke tailor that measures one client at a time, and only a half-dozen tech giants on earth can afford the fitting.
The company designs custom artificial intelligence processors alongside the high-speed Ethernet chips and optical lasers that wire thousands of them together. In the chip world, these custom parts are called application-specific integrated circuits, or XPUs. They are processors built strictly for one company's software rather than general use. Broadcom pairs that custom silicon with VMware, the corporate software engine that powers private corporate cloud networks.
The core business is printing cash. Third-quarter revenue jumped 86% to $29.6 billion, with dedicated AI chip sales surging 221% to $16.7 billion. Operating margins hit 67.9%, generating $13.7 billion of free cash flow in ninety days. Its contracted order backlog—what accountants call remaining performance obligations—stands at $179.2 billion. Management already secured the manufacturing capacity to double AI revenue to roughly $115 billion next year.
Yet the stock plunged roughly 14% over two sessions, sliding 38% below its peak. Supply-chain reports suggested Google trimmed one million of its next-generation custom processor orders to test rival designs from MediaTek. The equity now trades at roughly 15.7 times forward earnings. That forward price-to-earnings multiple measures how much investors pay today for every dollar of projected profit next year. Strangely, Broadcom's multiple is roughly double Nvidia's valuation.
Here is the real problem. A single distributor handles half of Broadcom's quarterly revenue, and its top five end clients account for 55% of total sales. That is not a diversified technology platform; it is a concentrated bet on a handful of hyperscale budgets. Because custom processors require heavy memory stacks, gross margins are already guided down toward 73% from 78%. If Google diversifies its orders, Broadcom loses its sole-source premium on the one partnership underpinning its entire growth narrative.
What we're watching: Fourth-quarter earnings on 9 December, arriving just hours after the Federal Reserve's rate call. We need AI semiconductor revenue to top the $21.7 billion guidance and gross margins to hold above 73%. Contracted backlog also has to expand beyond $179.2 billion, because that is what proves next year's $115 billion target is locked down.
Our take: Wait for a better entry. Generating 68% operating margins is the sign of an elite franchise, but buying at twice Nvidia's valuation multiple right before an unverified customer leak gets answered is an unnecessary risk.
2. Alphabet (GOOGL) — The Company Store
In an old mining town, the miners spent their wages at the company store because the owner controlled the whole valley. Alphabet sells ads, rents cloud computing, and now sells its custom processors directly to its corporate partners.
Most investors view Google as an advertising utility running Search and YouTube. That missed the quiet transformation inside Google Cloud, which now generates substantial product revenue by selling complete TPU computer systems. These Tensor Processing Units are Alphabet's proprietary artificial intelligence chips, designed in-house to train massive models without relying entirely on outside vendors.
The operational growth is undeniable. Third-quarter revenue rose 25% to $127.6 billion, with Google Cloud jumping 103% to $30.7 billion while delivering a 36.2% operating margin. An accounting writedown on outside startup investments obscured underlying operating profit growth of 30%. Total contracted cloud backlog reached $560.8 billion, with over half scheduled for delivery over the next twenty-four months.
Wall Street dumped the stock anyway, leaving it just above its 52-week low at 17.3 times forward earnings. Capital spending hit $51.4 billion in three months, pushing quarterly free cash flow into negative territory. Alphabet paused its share buyback program, issued convertible preferred stock, and saw its long-term borrowing climb above $125 billion. Investors fear the search giant is spending capital faster than its advertising core can generate it.
The bear case focuses on a tangled corporate loop. Alphabet owns roughly 13.4% of Anthropic, while Anthropic has committed hundreds of billions of dollars to cloud and chip suppliers over the coming decade. Skeptics argue Alphabet is essentially funding its own customer demand through venture investments. Meanwhile, enterprise surveys show rival labs taking corporate market share, and aggressive regulatory court orders now restrict exclusive search distribution deals. If capital spending crosses $300 billion without producing immediate cash flow, multiple compression will accelerate.
What we're watching: Fourth-quarter results in early February. The bull thesis requires full-year 2027 capital spending guided at or below $300 billion, separate disclosures on TPU hardware shipments, and cloud revenue crossing $38 billion with margins intact.
Our take: Accumulate. Alphabet is the only mega-cap cloud provider generating positive trailing cash flow while investing aggressively in silicon. Trading at a steep discount to the broader market, the shares already price in extreme regulatory and operational pessimism.
3. SoFi (SOFI) — The Bank That Is a Bond
A bank is simply a bond portfolio wearing a smartphone app. It lends money out for years at fixed returns while paying floating interest on customer deposits; when rates climb, the costume comes off.
SoFi holds a national bank charter wrapped in a modern consumer finance interface. It originates personal, student, and mortgage loans, gathers customer savings accounts, and rents its core payments engine to other financial institutions. Unlike traditional lenders relying on expensive Wall Street financing, SoFi deposits fund roughly 90% of its total loan book.
The latest operational results were stellar. Net revenue grew 44% to $1.385 billion, while adjusted EBITDA reached $469.8 million at a 34% margin. Deposits climbed past $51 billion, saving the firm roughly $745 million in annual interest expenses compared to wholesale credit facilities. Management lifted full-year revenue targets above $5 billion and projected roughly $0.90 in adjusted earnings per share for next year.
The stock collapsed anyway, sinking roughly 59% from its highs to hit fresh yearly lows. SoFi built its corporate budget expecting rate cuts, but the bond market shifted toward aggressive hikes instead. The stock now trades at 13.9 times next year's earnings guidance and just 1.67 times its tangible book value—the bare-bones liquidation value of its physical equity.
The catch sits inside the accounting marks. Loans held on balance sheets must be adjusted for fair value based on current interest rates. Rising discount rates triggered over $800 million in asset valuation losses last quarter, offset only by interest rate hedges that do not protect against credit defaults. While personal loan charge-offs sit at 2.26%, selling off delinquent loans masks the real loss rate, which sits near 3.4%. If the job market softens while rates march higher, those consumer balance sheets will crack.
What we're watching: The Federal Reserve's rate decision on 9 December, followed by fourth-quarter earnings in late January. We want all-in personal loan defaults capped near 3.4%, core equity ratios holding above 14%, and next year's $0.90 earnings target left firmly in place.
Our take: Accumulate. A digital bank compounding revenue at 44% with an insulated $51 billion deposit base is remarkably cheap at 1.67 times tangible book value. Treat it as a direct bet on the American consumer, and size your position accordingly.
4. Axon (AXON) — The Taser Subscription
Axon sells non-lethal weapons the way consumer brands sell razor blades, but its customer is city hall. The hardware gets officers equipped, while recurring software subscriptions to store digital video footage lock in police departments for half a decade.
The company provides TASER energy weapons, Body 5 cameras, in-car fleet surveillance, and cloud-based evidence management. It also sells Draft One, a machine-learning tool that transcribes body-camera audio into structured police incident reports, alongside Dedrone airspace security systems for federal installations. More than 95% of its revenue comes from recurring subscription packages.
Under the hood, software demand remains exceptionally durable. Third-quarter revenue rose 28% to $908.6 million, led by a 43% leap in software and cloud services. Net revenue retention hit 126%, meaning existing law enforcement agencies spent 26% more with Axon than they did a year ago. Future contracted bookings climbed 39% to $15.8 billion, supporting management's long-term goal of reaching $6 billion in revenue by 2028.
Despite that backlog, the stock sank roughly 52% from its late-summer peak to trade at 31 times forward earnings. Axon lowered its full-year revenue guidance by roughly $130 million after three major federal counter-drone deployments were pushed out amid federal funding and election cycles. Memory inflation added roughly half a point of margin pressure, adjusted gross margins fell to 59.2%, and nine-month free cash flow turned negative.
The short thesis is simple. Federal hardware orders are unpredictable, and TASER revenue, the franchise product, actually fell about 2% year over year. At the exact moment operating cash flow turned negative, Axon committed roughly $1.2 billion for a 37% stake in drone-maker Skydio, now settling about 90% in cash. Writing automated police reports also carries severe tail risk; if a machine-generated report invents facts in a criminal case, Axon faces unprecedented legal liability. Paying over thirty times earnings for a hardware-dependent business burning cash leaves no margin for error.
What we're watching: Full-year results arriving in February. Axon must demonstrate fourth-quarter free cash flow near $425 million to meet its annual targets, guide 2027 revenue growth above 30%, and push gross margins back across the 60% threshold.
Our take: Wait for a better entry. Axon owns an impenetrable public-safety software ecosystem, but burning cash while writing a $1.2 billion check is too risky when the equity still trades above thirty times forward earnings.
The bottom line
When the invoice arrives, you want to own businesses collecting real cash today rather than companies depending on cheap capital through 2028. Broadcom designs the critical custom chips but carries the customer concentration. Alphabet converts advertising cash into cloud infrastructure at a bargain multiple. SoFi funds rapid balance-sheet growth with sticky deposits, and Axon locks in municipal software revenue while working through federal hardware delays. We will track every one of these catalysts as the tape digests higher rates.
Analysis, not advice. We may own these names and we may sell them at any time.




