Dow DJI Prints 54,373.94 All-Time High as S&P 500 Stalls at 7,723.52 and Nasdaq Bleeds to 26,363.44 — Insmed INSM Rips 30%

The Dow printed 54,373.94 at 9:49 a.m. ET Thursday, a fresh all-time high, and by 9:57 it had rolled over to 54,296.73, down 52.39 points or 0.10%. That eight-minute round trip is the entire character of this tape. Wednesday's close of 54,349.06 marked the blue-chip index's third consecutive record finish and fifth straight positive session, built on a 263.18-point, 0.49% advance. The S&P 500 refused to come along. It lost 13.00 points, or 0.17%, to 7,723.52, snapping a four-session winning streak after tagging an intraday record of 7,793.68 in the morning — a reversal that left the index 70.16 points, or 0.90%, below its own high by the bell. The Nasdaq Composite got smoked, shedding 221.55 points, or 0.83%, to 26,363.44, its first down day after a run that carried it to 26,584.99 on Tuesday behind a 2.59% single-session rip.

Breadth confirmed the split. Decliners outnumbered advancers 1.15-to-1 on the NYSE and 1.34-to-1 on the Nasdaq. The S&P 500 logged 34 new 52-week highs against three new lows, while the Nasdaq Composite recorded 107 highs against 75 lows — a market where the winners are winning loudly and the losers are being taken out entirely. Volume ran 17.85 billion shares against the 17.34 billion 20-day average, so the rotation is happening on real size, not summer drift.

Thursday's pre-bell setup carried the same fingerprint. The Dow tracker ran +0.20%, the S&P 500 tracker +0.03%, the Nasdaq 100 tracker -0.88%, and the Russell 2000 tracker -0.09%. Dow futures were up 176.00 points to 54,670.00, S&P futures up 10.75 to 7,760.25, and Nasdaq futures down 132.75 to 29,482.25. The Russell 2000 had closed Wednesday at 3,019.19, off 0.59%, clinging to the 3,000 handle. Volatility stayed asleep: the VIX settled at 15.87, up 0.38%, then ticked to 15.93, up 0.76%, in the pre-market. Nobody is paying for protection into a payroll print.

Three sessions built this: Monday's 693.38-point Dow gain to 53,178.41, Tuesday's 907.47-point rip to 54,085.88, then Wednesday's 263.18. That is 1,170.65 Dow points, or 2.20%, in three days while the Nasdaq gave back 0.83% of it on the last one. The index that owns the AI trade is the one that stopped working.

The 199,000 Claims Print And A Labor Market That Won't Crack

Initial claims for state unemployment benefits rose 1,000 to a seasonally adjusted 199,000 for the week ended August 1, coming in 3,000 below the 202,000 consensus. That is the fourth straight week under 200,000 and it lands after a stretch that produced the lowest claim count in 57 years. Claims hit 187,000 in the week to July 18 — nearly a six-decade low — then rebounded 9,000 to 197,000 for the week ended July 25 before this week's essentially flat 1,000 tick higher. The four-week average had been running at 202,750.

Continuing claims tell a slightly different story. The number of people collecting benefits after an initial week of aid rose 24,000 to a seasonally adjusted 1.801 million for the week ended July 25, reversing a 7,000 decline in the prior period that had marked a one-month low at 1.782 million. That is the tension in this data set: almost nobody is getting fired, and almost nobody who loses a job is getting rehired quickly. The firing side of the equation has collapsed. The hiring side has not recovered.

Claims have dropped hard since surging in early June, and part of that decline is seasonal-adjustment noise around summer plant shutdowns. The more durable read is that layoffs stayed pinned at generational lows through an oil price shock delivered by a Middle East war now in its sixth month. Federal-employee claims, a category under scrutiny given the administration's workforce reductions, fell by 52 to 418 in the most recent detailed week.

The AI-displacement thesis has not shown up in the hard data. There is no evidence of widespread job losses tied to the artificial-intelligence buildout, and the cuts that have landed remain concentrated in technology. That distinction matters for how the tape trades software: the sector is compressing headcount and budgets simultaneously, which shows up in bookings and guidance before it ever shows up in a claims print.

The claims report carries no bearing on Friday's July employment report, which was compiled from a different reference period. Traders treated it accordingly — the 199,000 print moved nothing at the open. Rates held, the dollar held, and the Dow kept grinding at its record. A labor market this tight is not a reason to buy stocks; it is a reason to worry about what the Fed does next.

Challenger's 33,429 Is The Lowest July Since 2024

Planned job cuts by U.S.-based employers dropped 27% to 33,429 in July, the lowest monthly total since July 2024. Announced layoffs fell 46% from a year ago. Cuts announced so far in 2026 run 41% below the same period in 2025. The pace of layoffs collapsed through the summer, and the concentration is unambiguous: the announcements are coming out of technology, where AI investment is reshaping org charts rather than expanding them.

The hiring side of the same report printed a four-year high. Announced hiring plans jumped 47% in July to 16,095, though that increase came off a very low base — 16,095 announced hires against 33,429 announced cuts is still a ratio of roughly one to two. Year-to-date hiring announcements total 107,500, up 25% from the same stretch in 2025. The direction is right. The absolute level is not the kind of number that generates wage pressure.

Set the two series side by side and the low-hire, low-fire economy comes into focus. Job openings slipped to 7.36 million in June, down 178,000 from a downwardly revised May total and well below the 7.6 million consensus, with the openings rate easing to 4.4% of the labor force as healthcare-related postings drove most of the decline. That is a demand-side softening that shows up in vacancies before it shows up in separations.

The services side reinforced it. An industry survey released Wednesday showed the services-sector employment measure contracted in July, with respondents describing modest headcount reductions, some coinciding directly with AI implementation. So the anecdote is there even where the aggregate data has not caught it.

For equities, this configuration cuts both ways. A labor market that will not shed workers keeps consumption intact and keeps the earnings base from cracking — which is why the Dow keeps printing records on industrials, healthcare and financials. But it also strips away the case for policy easing at a moment when energy-driven inflation is still working through the system. The market is being asked to pay full multiples for earnings growth while the central bank has zero cover to cut. That is a narrower path than a 15.87 VIX implies.

The 33,429 figure is the cleanest single data point in the release. It says the corporate sector is not panicking. It does not say the corporate sector is expanding.

Friday's Payroll Print Is The Only Number That Matters

Nonfarm payrolls are forecast to have increased by 80,000 in July after rising 57,000 in June. The unemployment rate is expected to hold at 4.2%. Those two numbers now carry more weight for August positioning than anything in the earnings calendar, because the entire rate structure is sitting on the assumption that the labor market stays exactly where it is — soft enough to keep the Fed from hiking, firm enough to keep the earnings base intact.

The risk skews toward a higher jobless rate. A consumer survey released last week showed the share of respondents describing jobs as "plentiful" dropped in July to the lowest level since February 2021. That measure has historically led the unemployment rate by a quarter or two. If the participation rate holds and hiring stays at the 57,000-to-80,000 pace, the arithmetic pushes 4.2% toward 4.3% without a single additional layoff.

Job growth already decelerated hard from its spring pace. The three-month trend running through June sat well below the roughly 150,000 monthly rate that had defined the first quarter, and the June print of 57,000 was the weakest of the year. An 80,000 July would confirm a new, lower plateau rather than a bounce.

The market's problem is that a strong print and a weak print both create trouble. A number north of 120,000 with the unemployment rate steady at 4.2% hands the three hawkish dissenters on the policy committee live ammunition, and the 30-year yield — already at its highest since 2007 last week — has nowhere to go but up. A number under 40,000 with the rate jumping to 4.4% breaks the earnings-growth story that justifies an S&P 500 trading within 1% of an all-time high at these multiples.

That is why Thursday's session went nowhere despite a wall of individual-name violence. The Dow held its record, the S&P 500 ran +0.12%, the Nasdaq +0.17%, and the VIX stayed under 16. Positioning is frozen ahead of 8:30 a.m. Friday. The single-stock moves — a 30% ripper here, a 24% wipeout there — are earnings reactions being expressed inside a market that has stopped taking index-level risk. Nothing resolves until the payroll number lands.

Three Dissents And A Central Bank That Might Hike

The Fed left its benchmark overnight rate in a 3.50%-3.75% range last week on a 9-3 vote. All three dissenters preferred a quarter-point increase. That detail reframes every rate-sensitive trade on the board. This is not a central bank debating the timing of cuts. This is a central bank with a hawkish minority arguing for tightening into an oil shock, and a majority holding only because the labor market gives them room to wait.

The bond market got the message. The 30-year yield climbed to its highest level since 2007 in the sessions following the decision as traders repriced the hold as hawkish rather than neutral. The 10-year note traded down almost 7 basis points to 4.676% on Monday as oil slid on de-escalation headlines, then fell more than 6 basis points to 4.619% on Tuesday. The 2-year slipped more than 5 basis points to 4.198% in the same move. A 42-basis-point spread between the 2-year and 10-year with the policy rate at 3.50%-3.75% is a curve pricing neither recession nor easing — just a long hold at a restrictive level.

The inflation problem driving the dissents is energy, and energy is a function of the Strait of Hormuz. Six months of conflict have kept crude structurally elevated, and every headline that moves shipping capacity moves the front end of the inflation curve. Officials have said the effects of the conflict may require a response, though not immediately, with current policy positioned to allow a wait-and-see stance. The three dissenters disagree on the "not immediately" part.

For equities, this is the constraint that caps the multiple. A market trading at 7,723.52 on the S&P 500 after a 1.79% Tuesday and a record Tuesday close needs either earnings acceleration or a lower discount rate. Earnings acceleration is exactly what Thursday's tape shot down in software. And the discount rate is not going lower while a third of the committee wants it higher.

The stable labor market is what gives policymakers latitude to focus on the inflation fallout from the conflict rather than on employment risk. That latitude is the market's problem, not its comfort. Every 199,000 claims print and every 33,429 Challenger figure buys the hawks another meeting.

Fiserv Gets Smoked Again On A Guidance Cut Nobody Needed

Fiserv tumbled nearly 12% after slashing its full-year outlook for the second time in three quarters. Adjusted 2026 earnings per share guidance came down to $7.20-$7.40 from $8.00-$8.30, putting the new $7.30 midpoint 9.9% below the $8.10 consensus and 10.4% below the prior midpoint. Full-year organic revenue guidance was cut to a range of flat to down 1%, against a prior forecast of 1% to 3% growth.

The quarter behind it was worse than the guide. Revenue for the period ended June fell 4% year over year to $5.29 billion. Adjusted revenue came in at $4.96 billion, also down 4%, with organic revenue contracting 5%. Adjusted earnings per share dropped 26% to $1.84 from $2.47, missing the $1.91 consensus. Merchant solutions revenue declined 1%. Financial solutions revenue fell 8% — the segment that was supposed to be the stable half of the business.

Cash generation deteriorated in step. Net cash from operating activities came in at $2.08 billion for the first six months of 2026, down from $2.31 billion a year earlier. Cash and equivalents stood at $627 million as of June 30. For a business that carries meaningful leverage and was built on the promise of compounding free cash flow, a $230 million swing in first-half operating cash is not a rounding error.

The stock has been in free fall for eighteen months. Shares are down nearly 20% in 2026 after falling almost 68% in 2025, and the pre-market print at $49.54, off 8.45%, put the name at a level that would have been unthinkable when it was still valued as a dependable fintech compounder. An activist holder has escalated pressure for a full portfolio review and a board refresh following the rout. Leadership has churned, including a president's resignation in July on contractual grounds.

The pattern that matters for the sector: this is the third consecutive quarter where both segments contracted and management reset the bar lower. Organic revenue fell 4% in Q1 with adjusted EPS down 16%; Q2 delivered organic down 5% with EPS down 26%. The deceleration is accelerating. Payments volume growth exists — the problem is that it is no longer converting into revenue growth at this company, and the guide now embeds a full year of that.

HubSpot Prints Its Worst Day Ever Over 7,000 Customers

HubSpot plunged 23% by early trading, a move that would stand as the largest single-day decline in the company's 12-year public history if it holds into the close. The pre-market print of $192.91 sat 22.90% below Wednesday's $250.21 close and 63.3% below the 52-week high of $525.51 — closer to the $169.63 low than to anything resembling the old valuation.

The quarter beat. Adjusted earnings came in at $3.26 per share against a $3.02 consensus. Revenue hit $911.7 million, up 20% as reported and 17% in constant currency, ahead of the $898.13 million the Street modeled. Subscription revenue reached $894.0 million, up 20%. Non-GAAP operating income rose 44% to $185.3 million, lifting non-GAAP operating margin to 20.3% from 17.0%. GAAP operating income swung to a $43.3 million profit from a $24.6 million loss. Operating cash flow ran $222.8 million and the company repurchased $531.9 million of stock, with another $1.0 billion authorization added on August 3.

None of it mattered against the customer number. HubSpot added just 7,000 net new customers to reach 306,446, against an internal target of 9,000 to 10,000. Management then guided quarterly net additions down to roughly 5,000-6,000 for Q3 and beyond. Full-year revenue guidance was trimmed to $3.678-$3.686 billion from a prior $3.70 billion, and Q3 revenue was set at $924-$925 million — about 1.8% below consensus and implying deceleration to 14% reported growth from Q2's 20%.

The explanation is a deliberate pivot to trials and outcome-based AI pricing that extended sales cycles, colliding with a sudden tightening in software budgets that pushed more deals through C-suite and board review. Net revenue retention is guided flat at 102% for the year, with customer dollar retention holding in the high-80s and net upgrade rates expected to compress in the second half on budget optimization. Free cash flow is modeled near $750 million with capex at 5-6% of revenue.

There were bright spots buried in it. Deals above $120,000 in annual recurring revenue rose 38% year over year, 64% of new Pro Plus customers landed with multiple hubs, and Data Agent customers jumped 80% sequentially to 16,000. The tape ignored all of it. Multiple downgrades and target cuts landed by the open.

Datadog Beats, Raises, And Still Gets Sold 20%

Datadog tumbled about 20% in the pre-market and was down 16% by the open after beating on both lines and raising its full-year outlook again. That is the cleanest illustration of where the software bar now sits. The company delivered a Q2 non-GAAP earnings and revenue increase, lifted 2026 guidance, and lost a fifth of its market value on the print.

The setup explains the reaction. Consensus called for $0.58 per share on $1.08 billion in revenue, implying 30.22% top-line growth against the year-ago quarter, with full-year estimates at $2.41 per share on $4.34 billion — increases of 17.56% and 26.62% respectively. Into that print, price targets had been marched steadily higher across the sell side through early August, with one Street-high target at $330 built on customer and partner checks and a competitive position in observability spanning more than 33,000 customers. The stock had closed at $254.79 in a recent session after a 3.2% decline, so the multiple was already stretched into a set of expectations that had climbed for a month.

What got sold was the forward revenue guide, which came in softer than the raised full-year number implied on a quarterly basis, layered onto broader software-sector anxiety about AI monetization and visible insider selling pressure. The company had just posted its first billion-dollar revenue quarter, which raises the base and makes every subsequent growth comparison harder. Prior concerns had already surfaced that growth at current levels would be difficult to sustain into 2026 given comparisons and valuation, even with AI workloads contributing.

Stack it against HubSpot and the pattern is unmistakable. Both companies beat. Both companies guided forward revenue below what the Street wanted. Both lost roughly a fifth of their value in a single session. The AI-adjacent software complex is being repriced not on results but on the credibility of the growth curve twelve months out, and the market has decided it does not believe the curve.

That repricing is what dragged the Nasdaq Composite down 0.83% to 26,363.44 on Wednesday and kept the Nasdaq 100 tracker at -0.88% in Thursday's pre-market while the Dow tracker held +0.20%. The rotation is not a sector call. It is a duration call on software earnings.

Insmed Rips 30% As Brinsupri Clears $309 Million

Insmed ripped 30.18% to $128.90 in the pre-market, adding to a $21.46 billion market capitalization, after posting a quarter that reset the entire earnings model. Total revenue hit $425.5 million against a $393.74 million consensus, an 8.1% beat, with the year-over-year comparison approaching a quadruple as the commercial launch scaled.

BRINSUPRI generated $309.2 million in the quarter, up 49% sequentially — implying roughly $207 million in Q1 and a launch curve that is still steepening one year after approval for non-cystic fibrosis bronchiectasis. ARIKAYCE contributed $116.3 million, up 8% year over year, which keeps the legacy inhaled franchise growing rather than cannibalized. The mix shift matters for margins: the oral tablet formulation carries materially lower manufacturing costs than the older inhaled therapy.

The bottom line inflected violently. Net loss narrowed to $13.24 million, or $0.06 per share, from $321.69 million, or $1.70 per share, a year earlier. Consensus had modeled a $0.69 per-share loss. A $99.76 million gain from a change in fair value contributed to the swing, but the operating leverage underneath it is real — this is a company that was burning $300 million a quarter and is now within one strong quarter of breakeven, with cash-flow positivity targeted for 2027.

Management then raised the ceiling on the whole story. Peak revenue projections now run above $7 billion for BRINSUPRI and above $6 billion for TPIP, with ARIKAYCE reiterated above $1 billion — more than $14 billion in combined peak sales. Pre-clinical programs are guided to consume less than 20% of overall expenditures, with one to two investigational new drug applications per year.

The technicals were primed for this. Shares entered the print at a $104.32 50-day moving average against a $129.85 200-day, with a 52-week range of $90.39 to $212.75. The $128.90 pre-market print puts the stock 23.6% above its 50-day and right on the 200-day — the exact level that has capped every bounce since the spring drawdown. Every one of the 23 analysts covering the name carried a buy rating into the report with a $195.95 mean target, and insider selling had been running as a separate overhang.

Healthcare is where this market is paying. The S&P 500 healthcare sector closed up 1.3% Wednesday behind a 4.6% Amgen gain that added more than 100 Dow points on 9% sales growth, and a 4.9% Eli Lilly rip on $8.38 per share against a consensus $2.37 lower, with revenue up 47.7% to $22.97 billion versus $20.69 billion expected.

SiTime And SoundHound Show Where The AI Bid Still Pays

SiTime jumped 26.31% to $686.00 in the pre-market, a record level that blew through the prior 52-week high of $627.00, after revenue grew 127% year over year to $157.4 million from $69.5 million. That figure cleared the company's own $140-$150 million guidance range by $7.4 million at the top end and beat the $145.09 million consensus by 8.5%. Non-GAAP diluted earnings came in at $2.34 per share against a $1.93 consensus, with GAAP diluted EPS at $0.66.

The segment detail is where the AI infrastructure thesis lives. Every segment grew at least 50%, and the Communications, Enterprise and Data Center business expanded 181%. The mechanism is specific: newer inference infrastructure requires two to four times more timing content than AI training systems because synchronization requirements rise sharply when the objective is GPU utilization, latency reduction and throughput. Higher unit shipments, more timing content per system and rising average selling prices are all compounding at once, with premium products enriching the mix. Full-year revenue growth guidance had already been lifted to at least 80%, and the company cleared its antitrust waiting period on a $1.5 billion acquisition of timing assets, backed by a committed $900 million senior secured bridge facility.

SoundHound AI ran 25.97% to $8.10 from a $6.43 close on record quarterly revenue of $61.9 million, up 45% year over year and roughly ten times the Q2 2022 figure from its public debut. That beat the $52.4 million consensus by 18.1%. The adjusted loss came in at $0.02 per share against a $0.12 consensus loss. Adjusted EBITDA loss narrowed to $9.6 million from $14.3 million, improving margin to -15.5% from -33.5%. GAAP gross margin climbed 6.1 points to 45.1%, though GAAP net loss held at $42.8 million. Cash stood at $203 million.

Full-year revenue guidance was lifted at the low end to $230-$260 million from $225-$260 million, moving the midpoint to $245 million. With $106.1 million booked through June, the second half must deliver $123.9 million to $153.9 million. Regulatory clearances moved a pending acquisition closer to closing, with a combined-company 2027 revenue target of at least $350-$400 million. Short interest at 42% of the float turned the beat into a squeeze — the stock had bottomed at $5.65 and sat well below its $8.74 200-day exponential average going in.

The contrast with software is the whole story. Hardware and semis tied to AI inference are still being paid for beats. Application software is being sold on them.