A beat with two authors
Nothing in the headline is in dispute. Adjusted diluted EPS of $3.86 beat a $2.90 consensus (unverified — consensus aggregator), gross profit rose 18%, and the full-year EPS range moved to $12.20 to $12.70 from $11.00 to $11.50. The stock closed up 6.35% at $149.19 (unverified — market data). The chief executive, Jack Azagury, put it in one sentence in the release: "strength in cloud, infrastructure, and services, driven by AI demand."
That sentence names three things, and the filing shows they did very different jobs. Infrastructure wrote the revenue. Cloud and services wrote the profit. The AI demand is real; it shows up in hardware net sales, which rose 20.8% to $1.44B after 7% in the first quarter. It does not show up in the margin, and for a reseller the margin is the business.
The composite, and what it averages away
Eighteen percent gross profit growth on fifteen percent sales growth reads as a mix improvement. The composite hides the component. Product net sales rose $220.3M and product gross profit rose $8.9M, so product margin fell to 10.3% from 11.1%. Services gross profit rose $70.4M to $328.2M, a 63.9% margin. Services supplied 88.8% of the increase and now make up 62.9% of gross profit, against 58.3% a year ago.
The 10-Q says the same thing in North America's own terms: a 64 basis point product margin contraction from "a higher proportion of lower-margin sales," against an 87 basis point services expansion. The thesis-library rule applies to a reseller too: for the picks-and-shovels vendor, the funding question is a margin question. Nobody at Insight is borrowing to fund the AI server build. The party paying for the hardware growth is Insight's product margin. The chief financial officer, James Morgado, told us how much that line will earn for the rest of the year: hardware gross profit "will be up low single digits as component costs are impacting demand, particularly for devices" (unverified — call transcript). Read that against 20.8% revenue growth. Memory prices are lifting average selling prices and taking the margin out of them. The hardware boom is a revenue event with a gross profit rounding error attached.
The rest of this dive is for paid subscribers.
The headline numbers are above. The 7 sections below carry the mechanism, the peer read, the valuation work, and the dated tests that decide it.
- Who actually paid for the quarter
- The layer under the revenue line
- The raise is a first-half number
- The strongest case against me
- What decides it
- The trade (analysis, not advice)
- Bottom line
Already a subscriber? Sign in.
Who actually paid for the quarter
Cloud gross profit was $171M, up 39%, and the quarterly series ($123M, $130M, $138M, $139M, $171M) shows the step came in Q2. Agent services gross profit, booked at a 100% margin because Insight records only its fee, was $234M (+30%). Insight Core services, the work Insight delivers itself, was $95M (+21%). More than seven dollars in ten of services gross profit is a fee on someone else's product, and the largest someone else is Microsoft: about 32% of Insight's 2025 purchases, up from 27% in 2024.
Morgado's guide for the line is the tell: "We anticipate cloud gross profit to grow in the high teens to low 20% range as we move past the majority of the partner program changes we have previously discussed" (unverified — call transcript). Decode it with the 2025 filings. The Q2 2025 10-Q blamed "a decrease in certain fees from cloud solution offerings as a result of partner program changes." So the $123M base was a depressed base, and part of the 39% is a lap. The Q1 2026 10-Q then credited "improved performance in partner programs" in the product margin. The profit engine of this quarter is a fee schedule a single vendor rewrites, and the company is telling you growth will settle into the high teens once the easy comparison is gone. The 10-Q also says software and certain cloud sales are "typically seasonally higher in our second and fourth quarters." I'm not banking the step on one print: a mix-driven margin move is direction-agnostic — don't bank it until it prints twice, and the September quarter is the second print.
The layer under the revenue line
The balance sheet shows how big the agent business has become. The 10-Q is candid that netted costs, "while excluded from net sales and cost of goods sold, are processed and applied to accounts receivable and accounts payable," so the unadjusted cash conversion cycle does "not provide an accurate reflection of our cash conversion metric." Netted costs were $5.1B in the quarter, against $3.6B a year earlier and $2.5B two years earlier. Net sales over the same two years stayed between $2.00B and $2.40B.
That is why receivables rose $2.30B since and trade payables $2.26B, while trade working capital sits at $1.54B, down from $1.85B in March. The agent layer passes through the balance sheet almost gross in, gross out. It is not the risk here. The risk is timing: Insight pays partners on shorter terms than it collects, and in a June quarter the "timing of large partner payments" (unverified — call transcript) took operating cash flow to negative $12.2M.
The raise is a first-half number
Azagury was plain about it: "We always expected a stronger first half than second half and Q2 was always going to be a high watermark for us" (unverified — call transcript). The arithmetic agrees. First-half adjusted diluted EPS was $6.73, so the new range implies $5.47 to $5.97 for the second half, against $5.82 a year earlier: between 6.0% lower and 2.6% higher. Gross profit guidance of 8% to 10% growth on 2025's $1.76B implies second-half gross profit of $918.6M to $953.8M against $912.6M, growth of 0.6% to 4.5%, after 16% in the first half. The midpoint moved up $1.20; the two first-half beats against consensus add to $1.42 (unverified — consensus aggregator). The raise is smaller than the beat. Management banked the first half and guided the second flat.
Cash has the same shape. Holding the $300M to $400M operating cash flow range (unverified — call transcript) after $20.1M in the first half needs $280M to $380M in the second. Insight did $402.8M in the second half of 2025 and $339.8M in 2024, so the number is ordinary. It matters because the projected $299M of repurchases "would represent over 90% of our projected free cash flow" (unverified — call transcript). The buyback is sized to a second half that has to deliver.
The strongest case against me
The best argument on the other side is that I have the engine backwards. Hardware growth accelerated from 7% to 20.8% in one quarter, servers are "very, very strong" in Azagury's words (unverified — call transcript), and a reseller's product margin compresses early in a refresh and recovers as mix settles. Memory-driven price increases also lift the dollar value of every box. On that reading, the Q2 cloud step is seasonal, it fades in September, and product gross profit, off a Q3 2025 base of about $172M, grows faster than services gross profit, off $262M, for the first time this year. It is a real argument, and the seasonality part is right. I still come down on the other side because the company's own guide takes it apart: Morgado has hardware gross profit up low single digits for the year and says component costs are hurting device demand. Management does not expect hardware to become the profit engine, so the bull case needs the hardware line to beat its own guide by a wide margin.
What decides it
The third-quarter report, expected around (unverified — date not confirmed by the company). Two readings on one print. Cloud gross profit at or above $150M, about 15% over $130M, together with services gross profit growing faster than product gross profit, confirms the read: the fee layer is the engine and the partner-program lap was only part of the 39%. Product gross profit outgrowing services gross profit means hardware has become the engine and I'm wrong. Cloud below $150M with services still ahead is a partial result: the engine is right, but it was running on the lap. The secondary check comes with fourth-quarter results in : second-half operating cash flow of at least $280M, the floor the buyback plan assumes.
The trade (analysis, not advice)
At about $159.03, the stock is roughly 12.8 times the $12.45 midpoint (unverified — market price). That multiple is paying for a first half that already happened and a second half guided flat. The asymmetric question isn't hardware demand, which management has already discounted to low single digits of gross profit. It is whether the fee layer keeps compounding once the partner-program lap is gone. A September quarter with cloud at or above $150M would make the flat second-half guide look conservative going into the fourth quarter, when US enterprise clients "tend to spend more, particularly on product." A September quarter with cloud stalling near the $130M base would leave the stock priced on the June quarter. Anyone who wants the AI server cycle should look at the component vendors; buying Insight means buying Microsoft's partner economics, with a hardware reseller attached.
Bottom line
Insight's strongest quarter of the year came from two places that barely touch. Hardware brought in the revenue and very little profit. Cloud and agent fees, built on a Microsoft program Insight does not control, earned most of the gross profit. The guidance raise banks the first half and guides the second flat. My view is that the fee layer, not the server cycle, is what Insight's earnings depend on. The third-quarter report settles it: services gross profit growing faster than product gross profit, with cloud at or above $150M.