|
Good morning. TSMC expects another huge quarter. Investors are still selling the companies building around its chips. That is the tension this morning: demand for computing looks strong, while the cost of financing it is getting harder to ignore.
CoreWeave lost 15% yesterday, Netflix is cutting roughly 1,000 jobs, and federal regulators want more answers from OpenAI and Anthropic. We also have Nvidia’s latest push into financial data, fresh inflation context and the stocks that held up best in Wednesday’s selloff.
BEFORE THE BELL
December contracts · October 15, 6:46 a.m. ET · Delayed Yahoo Finance quotes; change versus prior close.
|
|
01 / EARNINGS · TSM
TSMC sees a $53 billion quarter ahead.
|
|
|
TSMC R&D center, archive. 曾 成訓 / Wikimedia · CC BY 2.0.
TSMC reported $46.72 billion in third-quarter revenue, up 41.2% from a year earlier, and earned $4.60 per U.S.-listed ADR. Revenue cleared the top of its own $44.6 billion–$45.8 billion forecast. The next quarter looks bigger still.
Fourth-quarter revenue guidance is $52.3 billion–$54.3 billion. The $53.3 billion midpoint would mean another 14% increase from the quarter just reported. That is a substantial sequential jump for a business already operating at this scale.
Premarket: TSM traded at $421.00 (-2.09%) at 6:56 a.m. ET. Yesterday it closed down 3.40%. Strong operating numbers are arriving in a market that is marking down the wider AI trade.
Profitability remains striking. Gross margin reached 66.4%, and operating margin was 58.7%. For the fourth quarter, management expects gross margin of 65%–67% and operating margin of 56.5%–58.5%. The revenue forecast points higher; the margin forecast does not promise another step up.
The technology mix helps explain the business. Two-nanometer chips supplied 8% of wafer revenue, while 3nm contributed 33%. Together, technologies at 7nm and below accounted for 79%. These percentages describe wafer revenue, not a count of chips shipped.
For investors in Nvidia, AMD and other chip designers, this is evidence that advanced manufacturing demand is reaching the factory floor. TSMC gets paid to make the chips. Its customers then need to sell their products, and the companies buying those products need to earn an acceptable return from using them. Those are related businesses with different economics.
That distinction matters after yesterday’s selling. Healthy orders at a manufacturer can coexist with pressure on a heavily financed data-center operator. One measures production demand; the other also has to contend with electricity, construction, borrowing costs and the timing of customer payments.
The next useful comparison is the one management has already supplied: can TSMC deliver that fourth-quarter sales increase while keeping margins within its forecast? More revenue is valuable, but the profit left after expanding capacity determines how much of the growth reaches shareholders. The results page below puts the release and presentation together so readers can compare the quarter with the outlook.
|
|
02 / STOCKS · AI INFRASTRUCTURE
CoreWeave lost 15%. Credit markets noticed.
|
|
|
Oracle’s Austin campus, 2018 archive. DronePhotographer / Wikimedia · CC0.
CoreWeave fell 15.46% to $65.78 Wednesday. Nebius dropped 10.08%, Oracle lost 8.83%, and Dell sank 11.27%. This was a broad hit to the companies supplying and operating AI infrastructure, rather than one disappointing earnings release.
This morning brought a second warning from the debt market. Reuters reported that Oracle’s five-year credit-default-swap spread widened 26 basis points to a record 288. CoreWeave’s spread moved above 1,000 basis points. These contracts price protection against a credit event; wider spreads mean that protection has become more expensive.
Reuters also reported CoreWeave’s 2032 debt trading at about 77 cents on the dollar. That adds context to the stock decline: investors are demanding a larger discount to own its existing debt as well. A CDS spread is not the company’s borrowing rate or a direct forecast of its probability of default.
The underlying funding problem is easy to understand. A data center absorbs money before it generates revenue. Buildings, servers and power connections have to be ready before a customer can use the capacity. Interest payments do not politely wait for that process to finish.
If lenders become more cautious, the next expansion can cost more to finance. That can reduce the return on a project even when customers still want the computing power. It can also make refinancing less comfortable for companies whose investment plans depend on continued access to capital.
For shareholders, the useful distinction is between contracted revenue and cash available today. A multiyear customer commitment can support a project without covering every near-term construction bill. Payment schedules, financing terms and the date a facility starts operating all affect how far that commitment goes.
TSMC’s results offer strong evidence of demand upstream. The credit market is asking a separate question downstream: who can afford to keep building? The two signals can point in different directions without either being wrong. Today’s credit-spread report gives a clearer view of that concern than a list of falling share prices alone.
|
|
03 / STOCKS · NFLX
Netflix cuts 1,000 jobs.
|
|
|
Netflix offices, 2019 archive. Coolcaesar / Wikimedia · CC BY-SA 4.0.
Netflix is eliminating roughly 6% of its workforce, or about 1,000 positions. Product, technology, legal and human-resources teams face the largest impact, according to the memo co-CEOs Greg Peters and Ted Sarandos sent employees Wednesday.
The stock finished at $70.57, up 0.16%, holding up considerably better than the AI infrastructure names. Management said the business remains strong and growing, and that Netflix is still on track to meet its full-year forecast.
The memo frames the cuts as a decision about where to put people and resources while continuing to invest in growth. That explanation leaves investors with a practical question: how much spending disappears, and how much simply moves to another part of the company?
A 6% reduction in employees does not imply a 6% reduction in total expenses. Compensation varies by role, some departures can bring severance costs, and content and other operating expenses remain. The announcement does not provide enough detail to turn the headcount figure into a reliable earnings-per-share estimate.
There is also an execution tradeoff. Product and technology teams help determine how well the service works and how easily customers find something worth watching. Reducing their size only improves the business if Netflix can maintain that experience while spending less or redirecting work more effectively.
For the next results discussion, the useful follow-up is specific: any restructuring charge, the timing of savings and whether the operating-margin outlook changes. Investors do not need another broad promise of discipline. They need to see what that discipline does to the accounts. Variety has published the staff memo in full at the link below.
|
|
04 / DEALS · PAG
Penske shareholders get a $212 cash offer.
|
|
|
Sytner BMW dealership, Newport, 2009 archive. John Grayson / Geograph · CC BY-SA 2.0.
Penske Automotive has agreed to go private at $212 per share in cash. The buyer is owned by Penske Corporation and Mitsui, already the company’s largest shareholders. The deal puts an enterprise value of about $16.3 billion on the business, including debt.
The offer is 8.84% above Wednesday’s closing price, according to the company. This follows the earlier approach in July; today’s development is a signed merger agreement. For shareholders, the question shifts from what the dealership business might earn to when the cash arrives.
Closing is expected in the first quarter of 2027. Regulatory and shareholder approvals are still required, including approval from unaffiliated shareholders. There is no financing condition, removing one potential escape route for the buyers, but the transaction is not complete.
Any gap between the stock price and $212 therefore reflects time and deal risk, rather than a guaranteed return. Penske also expects to keep paying regular quarterly dividends before closing, within the agreement’s limits.
The next scheduled update is October 28, when Penske will release third-quarter results. It has canceled the accompanying earnings call because of the transaction. The announcement below spells out the payment, approvals and dividend terms.
|
|
.jpg){kind=link}
