SK Hynix’s Selloff Is Really a Test of HBM Durability

Zscaler’s 8.1% weekly decline was not triggered by a guidance cut. It reflected a harder valuation question: what multiple should investors pay when a cybersecurity company moves from roughly 25% growth toward the high teens?

The company ended fiscal 2026 with annual recurring revenue of about $3.8 billion, up 25%, and a record 24% non-GAAP operating margin, according to the source. At its October investor day, Zscaler reaffirmed fiscal 2027 guidance for revenue of $3.908 billion to $3.938 billion, representing growth of 16.6% to 17.5%. ARR guidance was $4.396 billion to $4.426 billion.

That deceleration does not mean the business is weak. Zscaler’s zero-trust platform remains relevant as companies replace network-perimeter security with controls that connect users directly to applications. The company’s research unit also reported rapid growth in ransomware data theft, supporting continued security spending.

The valuation has already reset. TIKR placed the stock near 42 times forward earnings, down sharply from triple-digit multiples several years earlier. Its model assumed about 16.7% revenue growth, a 24.9% operating margin and a similar exit multiple. The projected upside therefore depends less on multiple expansion than on management reaching the modeled margin.

Cash flow is the complication. The source said fourth-quarter free cash flow fell to 7% of revenue as capital spending nearly tripled to $219 million. One quarter of elevated investment does not define the trend, but investors need evidence that free-cash-flow margin can recover toward management’s longer-term targets.

Zscaler’s next phase is about quality of growth. If recurring revenue expands near guidance while margins and cash conversion improve, the lower multiple can support attractive returns. If growth slows without corresponding operating leverage, a 42-times earnings multiple may still be demanding.

The next reports should be read across four measures together: ARR growth, large-customer additions, non-GAAP operating margin and free-cash-flow margin. Any valuation argument that ignores one of those four is incomplete.

Robinhood’s second quarter was strong enough to set records, yet the stock’s valuation still requires more than stability. Revenue rose 32% year over year to $1.31 billion, diluted earnings per share increased 48% to $0.62 and net deposits reached roughly $22 billion.

Trefis tested what would happen if the quarter simply repeated for a full year. Multiplying quarterly profit by four produced about $2.24 billion of annual earnings. That is a scenario, not company guidance, and it ignores seasonality and changes in trading activity. Even under that assumption, the source estimated a valuation near 39.6 times earnings, well above the S&P 500 comparison it used.

The exercise is useful because Robinhood’s revenue responds to customer activity, asset prices, interest rates and trading volumes. A flat quarter can still be profitable, but a premium multiple normally reflects continued gains in funded accounts, assets, deposits and higher-value products.

The operating data remain encouraging. Robinhood reported 4.8 million Gold subscribers in the second quarter. July net deposits were later finalized at $5.6 billion, and the company said trailing 12-month deposits represented 25% growth relative to the prior-year asset base. Management also lowered and tightened its 2026 adjusted operating-expense and stock-compensation outlook.

The downside is that a high multiple can contract before earnings deteriorate. Robinhood experienced that dynamic during the 2022 market shock, when revenue declined and the company posted an operating loss. Today’s business is broader and more profitable, but trading-sensitive income still introduces cyclicality.

The next report should be judged on net deposits, platform assets, subscriber growth, transaction revenue mix and expense control. Investors should also separate reported results from annualized scenarios: repeating one record quarter does not prove a durable run rate.

Robinhood does not need growth to avoid losses. It needs growth to defend the price investors are paying. If deposits and product adoption remain strong, the premium can persist. If they stall, valuation compression may matter more than near-term profitability.

Ondas shares have fallen despite a rapid series of defense orders, acquisitions and a new artificial-intelligence venture. That tension has encouraged retail traders to frame the stock as a possible short squeeze. The business evidence points to a more useful question: can Ondas convert its order announcements into revenue and cash?

The company said it secured more than $165 million of orders from the second half of August through October 1. Combined with roughly $105 million disclosed earlier in the third quarter, orders since June 30 exceeded $270 million. Ondas also formed Regna AI with Palantir, a standalone venture intended to bring specialized AI models into sovereign and defense environments.

Those developments expand the opportunity set, but orders are not the same as recognized revenue. Delivery schedules, acceptance milestones, production capacity and customer funding determine when booked demand reaches the income statement. Defense contracts can also be lumpy and subject to changing procurement timetables.

The source reported short interest above 40% and quoted comparisons with GameStop. That is sentiment, not a fundamental catalyst. A high short-interest ratio can amplify price moves in either direction, but it does not establish the value of Ondas’s order book or guarantee a squeeze.

Capital structure is another necessary part of the analysis. Ondas completed a large registered direct offering in January 2026, generating approximately $959 million of net proceeds. That improves funding capacity, but the issuance also increased the share base. Investors should evaluate growth per share, not just aggregate orders.

The next decisive evidence will be backlog conversion, gross margin, operating cash flow and updated delivery schedules. Regna AI will also need named customers or disclosed economics before it can be valued as more than an early strategic initiative.

Ondas has created a stronger operational story than the recent share decline suggests. Still, the route to durable equity value runs through execution and capital discipline, not social-media enthusiasm or the percentage of shares sold short.

CoreWeave’s backlog is large enough to support years of growth, yet the cost of building that capacity is becoming the central investment risk. HSBC maintained a Reduce rating and a $39 price target, which implied about 36.5% downside from the stock’s prior close, according to Stocktwits.

The bank’s concern is not a lack of customers. CoreWeave reported approximately $104 billion of revenue backlog at June 30. The company also signed large contracts during 2026 and has relationships across the AI ecosystem. The question is whether those contracts can generate returns above a financing cost that has risen sharply.

HSBC estimated that CoreWeave could produce about $12 million of annual revenue per megawatt of capacity in 2027, compared with $17 million to $18 million for larger cloud providers. It also modeled long-term operating margins of 15% to 16%, below management’s cited target near 25%. Those are bank estimates, not company guidance, but they explain why the stock can fall even while backlog expands.

Debt markets add pressure. The source said CoreWeave bond yields were near 15%. The company’s April offering included $1 billion of 9.75% senior notes due in 2031, and it later announced a $3.5 billion senior-note offering. High coupons can consume a meaningful share of project economics before shareholders receive any benefit.

CoreWeave’s second-quarter release showed both sides of the case. Backlog rose to $104 billion, giving strong visibility, while the balance sheet included billions of dollars of current recourse and non-recourse debt. Contracted revenue is valuable, but it arrives over time; interest and construction spending arrive sooner.

The next test is not whether AI demand exists. It is whether new capacity comes online on budget, revenue per megawatt holds up and operating cash flow begins to cover the capital structure. If those measures improve, the backlog can justify the risk. If they do not, more growth may increase financing pressure rather than create equity value.

Oracle’s $638 billion backlog looks like extraordinary revenue visibility. It also raises a less comfortable question: how much of that demand depends on one customer’s ability to keep funding a historic computing expansion?

Investor Steve Eisman said in an October 16 interview that roughly half of Oracle’s backlog is tied to OpenAI. That would equal about $319 billion. Oracle has not publicly broken out the customer’s share, so the figure should be treated as Eisman’s estimate, not as company disclosure.

Oracle’s reported numbers establish the scale. The company ended fiscal 2026 with $638 billion of remaining performance obligations, up from $553 billion one quarter earlier. Fiscal-year revenue was about $67.4 billion. In other words, contracted business now spans many years of current revenue, and the pace at which it becomes recognized sales matters as much as the headline amount.

The concentration question is sharper because Oracle is building capacity before receiving all of that revenue. The source cited negative free cash flow of $23.7 billion for fiscal 2026 and total debt of about $167 billion. Those figures mean the company is financing a substantial portion of the buildout while customers scale their own economics.

A large backlog is valuable only if contracts are enforceable, customers can pay, and infrastructure becomes available on schedule. Customer concentration increases the damage from renegotiation, delayed deployment or weaker demand, even if the total order book remains unchanged.

There is a bullish interpretation. OpenAI and other model developers need scarce computing capacity, Oracle’s cloud infrastructure revenue rose 93% in its latest reported quarter, and the backlog may support years of growth. The bear case is that one customer represents too much of the promised demand while Oracle bears capital costs now.

Investors should resist turning Eisman’s estimate into a settled fact. The better test is observable: watch how quickly remaining obligations convert into revenue, whether capital expenditure produces improving free cash flow, and whether Oracle discloses broader customer diversification. Until then, the backlog is both the company’s strongest asset and its most important concentration risk.