Nvidia Still Wants a New High; Salesforce Jumps 12% After Hours
US Stocks · Options · News · Views
The Indexes Went Flat; PCE Did Not Rewrite September
Wednesday, August 26, the cash session barely moved. The Dow fell 0.21% to 53,464. SPX was almost unchanged at 7,676. Nasdaq slipped 0.06% to 26,136. The cash tape handed the session to after-hours prints. NVDA cleared space during the day and got priced after the close. CRM and a cybersecurity name also waited until after the bell.
July PCE did not help. Core rose 0.2% month over month, in line, and is still 3.3% year over year. Headline rose 0.2% and 3.7%, a touch hotter than 0.1% and 3.6%. Two-year yields ticked up after the print, then faded. Money markets still write September as about a 60% chance of an unchanged rate. The 10-year closed around 4.66%, versus about 4.64% the prior day. The long end did not ease on this inflation print. Treasury buybacks of long bonds remain a signal. One session does not prove the suppression worked.
The second-quarter real GDP annualized revision stayed at 1.5%, in line. Consumption, exports, and investment carried the print. Government spending fell and offset part of the gain. Both macro releases cleared the line and gave no tailwind. The real pricing was three after-hours items: the META settlement, the CRM print, and the NVDA call.
Meta Settled; The Check Is Spread Over Ten Years
The state cases had been talked about in the $1.4 trillion neighborhood, up against META’s own roughly $1.4 trillion market cap, as if a loss could wipe the company out. That was never the point. The questions were the size of the penalty and whether more of the same would follow. Wednesday’s settlement took that noise off the table.
The figure is up to about $18 billion. META denied the claims that product design induced addiction in minors and that it improperly collected children’s data. It still agreed to cap daily use for minors, with mandatory pauses, nighttime blocks, stronger age checks, parental controls, and tighter content filters. More than 70%, about $12.7 billion, is guaranteed over ten years — a little more than a billion a year. The remaining 30%, about $5.3 billion, is contingent: platforms such as TikTok and YouTube have to adopt similar limits before that slice is paid in full. One social platform does not take the whole bill.
After hours the stock spiked, then settled around a 1% gain. The high was about 590, still inside 581 to 632. With the legal noise stripped out, the stock has to go back to what actually caps it: AI monetization is the slowest among the large platforms. It looks like the cheapest of the four megacaps for a reason. Give it about six months. If monetization shows up, the box can open. If it does not, spending has to come down, closer to Apple’s lighter-spend path. 581 to 632 is a box, not a direction.
Salesforce Up About 12%; 235 Did Not Clear
CRM jumped more than 10% after hours, as much as about 14%, and later held near 12%. Second-quarter adjusted EPS was $5.90, well above about $3.27. Revenue was $11.35 billion, up 11% year over year, a touch above about $11.33 billion. Gross margin was 77%, flat with last quarter and a bit under 78% a year ago. Growth has slowed from a sustained 20%+ before 2022 to about 10% in recent years. That is why the stock has been under pressure. The market story was that AI agents would replace high-margin software subscriptions.
Inside $5.90 there is a block of strategic-investment gains. Operating profit is not as explosive as the headline. What reset the mood is that the company answered the “will we get kicked out” question more clearly. It expanded the Anthropic partnership under the name Claudeforce: CRM’s customer system plugs into Claude, so users and agents can reach data and functions through a plugin. Agentforce already let companies deploy agents. AI and data products already throw off nearly $4 billion of recurring revenue a year. This time the interface is more direct. The claim is that AI is a growth tool inside the franchise, not a substitute for it. In software, the two names that worried people most were CRM and ADBE. If their own tools or third-party tools lift revenue, the market pays them. If not, it punishes them.
Fiscal 2027 adjusted EPS is now $16.67 to $16.71, above the prior $14.06 to $14.12. Revenue is $46.1 to $46.4 billion, about $200 million above the old guide. The quarter beat. The guide beat. That is a different posture from Tuesday’s INTU, which cut price to defend share. CRM wired AI into its own stack and lifted the outlook. The tape treated it as a large positive.
For valuation, stick to the current year for now. The updated January 2027 forward band is 332 to 498, cheap. The January 2028 set gets a question mark: this year’s raise already overshoots what the old model had for the following year, and pushing the same logic forward draws a down year that does not parse. The accounting still needs more data. After hours the stock tagged the lower end of 235 to 277. The high was 235. The heavy shelf is still there. It has not cleared, so this is not yet a right-side break. 277 is a long way up. A one-tape smash through is a high bar. The fundamental trend is turning better. Long-term holders can give it more time. Near term, watch whether 235 first kicks the stock back before it tries to grind through.
Nvidia Dumped Then Recovered; 236 Is Still the High
NVDA’s quarter was solid. Next quarter’s gross-margin guide came in a bit light. The stock fell about 3%, then rose about 4%. The second leg started after the call, when management talked demand hotter.
Second-quarter adjusted EPS was $2.22 versus about $2.09. Revenue was $96.22 billion versus about $92.38 billion. Data Center was $89.02 billion versus about $85.86 billion. Hyperscale data center was $48.71 billion versus about $43.55 billion. AI cloud, industrial, and enterprise was $40.31 billion, below about $41.96 billion. Edge Computing was $7.2 billion versus about $6.61 billion. Compute and networking was $88.3 billion versus about $84.69 billion. Adjusted gross margin was 75.0%, in line. Adjusted operating expenses were $8.23 billion, under about $8.32 billion. R&D was $7.05 billion, a touch above about $7.00 billion. Adjusted operating income was $63.96 billion versus about $61.19 billion. Most of the quarter cleared.
$96.2 billion is a huge print. Close to half of the business still sits with hyperscale, cloud, industrial, and enterprise customers. Those customers often cannot cover the purchases with their own free cash flow. Some of the money is outside financing. Some is cash or guarantees from Nvidia. That circular credit is one reason the backlog keeps filling. Demand is still growing. Cash is still coming in. The landmine has not gone off. It goes off when a link in the chain actually fails. For now the game is still on.
Third-quarter revenue is $108.0 billion plus or minus 2%, above about $104 billion, still with no China Data Center compute in the outlook. Gross margin is 73.5% to 74.5%, under the 75% that was expected. The guide is still well above the sell side. For some buyers it is not enough: some had $109 billion. Management said demand is still accelerating at this scale, and that fiscal 2028 revenue will grow about 70%, above the prior 40% to 50%. Amazon committed to use materially more Nvidia product. The margin swing is memory. This guide already assumes the company raises price. If memory keeps ripping, Nvidia either cuts bits or gives more margin. For memory names, a fiscal 2028 revenue growth rate near 70% is at least a narrative support. If algorithms or architecture cut memory intensity, that is a different story.
On the high-growth EPS path — about 87% this year, about 41% next, about 58% five-year compound — 30 to 40 times puts January 2027 at 267 to 356, midpoint 312, cheap; January 2028 at 377 to 503, midpoint 440. A $5 trillion company on a more conservative 20 to 30 times puts this year at 178 to 267. Taking 209 down through 180 only works if demand breaks. After hours it reclaimed 218, a small upgrade from 216. From a 218 open, overhead is thin. 236 has been the high for a long time. Recent tests tagged 227 and 230 and failed. After this gap higher, monthly, weekly, and daily technicals do not argue for an instant fade. The window is open. The bid still has to show up. The market cap is huge. If sellers do not step aside, the stock does not go.
The position stays. No trade. A move back above 260, even 300, is a long-term path, not a few sessions. First 218 has to hold. Then this tape has to see whether 236 finally prints. The semiconductor story can still be told. The circular-credit charge is still buried behind it.