The setup
Six companies that sit between the chipmakers and the people who run the equipment reported inside one week: Super Micro on , CDW and Avnet on , and Arrow, Insight and WT Microelectronics on . Every one of them sold far more hardware than a year ago. Every one kept less gross profit on the new dollar than its old book earned, Super Micro's year-ago comparison aside, and every one leaned on somebody else's balance sheet to do it. At midday on Arrow was down 9.4% and Insight 7.7% on prints that beat, and Avnet was down 7.6% the day after its own (unverified, market data, not closes).
So is this the norm for a distributor, or something worse? We pulled the filed history back through the last cycles to check. The answer is less alarming than the cash lines read and less comfortable than the income statements do. Burning cash on the way up is what these businesses have always done.
The tell is one line, and it is not the burn. Avnet's payables are 95.9% of its inventory, against a range of 54% to 85% from 2017 through . WT's payable days are 103, against 62 to 83 before this quarter. The shape of the cycle is normal. How much of it rests on supplier credit, at gross margins at or near the lows of the series, is not.
Our call: Avnet and Arrow together end the December quarter with at least $15.48B of inventory, the $8,175.1M and $7,306.0M they hold today. Below $15.48B, the parts shipped faster than they were replaced, and we are wrong.
Update, . Ingram Micro was added after this piece was emailed. It reported on , a week before these six, and its own page has the full read. The six-name argument and both calls are unchanged.
The numbers
| Company | Sales Y/Y | Gross margin | Gross profit per new sales dollar, Y/Y | Operating margin |
|---|---|---|---|---|
| CDW | +17.7% | 20.5% (21.9%) | 12.8¢ | 7.2% (7.7%) |
| Insight | +21.8% | 21.4% (21.7%) | 20.3¢ | 2.8% (4.6%) |
| Arrow | +39.0% | 10.8% (10.8%) | 10.8¢ | 4.0% (2.3%) |
| Ingram Micro | +20.8% | 6.55% (6.90%) | 4.9¢ | 1.93% (1.77%) |
| Avnet | +81.7% | 10.0% (10.4%) | 9.5¢ | 4.4% (2.4%) |
| WT Micro | +81.6% | 3.60% (3.74%) | 3.4¢ | 2.38% (1.72%) |
| Super Micro | +297.5% | 12.7% (9.3%) | 13.9¢ | 10.4% (3.6%) |
Source: company filings and WT's investor deck, to 5, 2026. Margins are not comparable across the group because of netted revenue and mix. Avnet's quarter had 14 weeks against 13; without the extra week its sales rose 70.6%. Notes ↓
Who paid for the build
Payables rising beside inventory does not mean suppliers funded inventory, because a distributor running other people's goods through its ledger carries a receivable and a payable for the same transaction. Arrow is the clean example. Its payables rose $5,754.5M in the quarter against $1,366.4M of inventory, which on the simple ratio is 421.1%, and that is the wrong answer to the right question. Net of receivables, suppliers covered 36.9%.
| Distributor | Inventory rise | Gross share | Net of receivables | Who carried the rest |
|---|---|---|---|---|
| WT Micro (NT$B) | 190.44 | 98.6% | 79.0% | A GDR and convertible raise of about NT$29.6B (unverified, press reports) |
| Avnet ($M) | 2,105.7 | 84.9% | 43.4% | Banks and bondholders: debt rose $955.1M |
| Ingram Micro ($M) | 546.6 | 230.2% | 74.4% | Vendors, mostly: debt rose $105.2M |
| Arrow ($M) | 1,366.4 | 421.1% | 36.9% | Banks: debt rose $581.4M through a securitization enlarged first |
Three distributors, one direction, very different exposure. The net share is our construct and no company reports it. Arrow's filing does not split supply chain services balances, so that number carries a band of 15.3% to 58.5% on its own page.
Ingram Micro is the variation. It produced $251.4M of operating cash in a quarter when Avnet and Arrow burned it (WT has not filed cash flow yet), and the reason is on its own release, which says it financed growth significantly with vendor terms. Payables rose $1.26B and payable days went to 83.5 from 74.5. Its inventory rose 9.1%, a far smaller build than Arrow's or Avnet's, and the nine months still used $1.26B. So the cash is the same supplier credit arriving by a different door.
The other three found different payers. Super Micro's customers funded the ramp: deferred revenue rose $3,054M, and without it operating cash flow would have been −$3,488M rather than −$434M. CDW has an intermediary holding inventory for one customer contract, $222.4M of net financing that sits inside a nine-month adjusted free cash flow up 17.5%; before that line the measure fell 15.8%. Insight spent $225.0M on buybacks against $127.5M of nine-month free cash flow, and its asset-based revolver covered the difference. Nobody in the group paid for this entirely out of earnings.
Is this the norm?
Mostly, and the history is unusually clear about it. Avnet posted negative operating cash flow in fiscal 2010, 2017, 2022, 2023 and 2026, then took in $690.0M in fiscal 2024 and $724.5M in fiscal 2025 as sales fell. It took in $1,118M in fiscal 2009. Arrow took in $850M in 2009 and $1,360M in 2020. A distributor's cash runs against its cycle, because working capital absorbs growth and releases it in decline.
| Company | Inventory ÷ quarterly sales | History | Gross margin | History | Read |
|---|---|---|---|---|---|
| Avnet | 76.3% | 56% to 102%, since 2017 | 10.0% | 10.4% or higher, since 2009 | Ratio mid-range, margin a low |
| Arrow | 68.2% | 35% to 73%, since 2010 | 10.8% | 10.8% or higher, since 2010 | High, not a record; margin ties the low |
| WT Micro | 85.2% | 53.4% to 75.9%, prior five quarter-ends | 3.60% | 3.41% to 4.58%, since 4Q24 | Out of range, on a short history |
| Ingram Micro | 43.1% | 41.4% in June, 42.6% a year ago | 6.55% | 6.50% to 6.90%, seven quarters | Normal, on a short history |
| CDW | 15.1% | 8.8% to 17.7%, since 2011 | 20.5% | 15.5% to 21.9%, since 2011 | Normal |
| Insight | 10.9% | 6.7% to 14.5%, since 2010 | 21.4% | 12.6% to 21.7%, since 2010 | Normal |
| Super Micro | 75.5% | 63% to 120%, since 2011 | 12.7% | 6.3% to 18.8%, since 2011 | Low only because sales nearly doubled |
Avnet's −$842.4M is a record in dollars and not as a share of sales: −7.9% against −9.6% in . Arrow's −$441.4M is −4.1% of sales, against −5.5% in . Its $7.31B of inventory is 25.8% above the $5.81B peak of . Avnet's $8.18B is 33.7% above the $6,116.0M peak of , on sales 1.59 times the old peak. New levels, and ordinary ratios.
Here is what is not ordinary. Gross margin sits at or near the low of the series at both US distributors, so the same write-down costs more of the profit. Avnet's payables reached 99.8% of inventory in June and 95.9% now. WT's payables are 72% of total assets and 4.99 times equity, and on the company's definition 103 days is 20 days above its previous high. Credit that large can be withdrawn by people who do not report to the distributor, and history gives us no reading at this level to compare it with.
Then there is the part of the inventory that is price. Jacobson, Avnet's chief financial officer, put memory pricing at about a third of Avnet's inventory increase (unverified, third-party transcript). Arrow's chief financial officer said most of Components' growth in the quarter came from price (unverified, third-party transcript), and Components kept 3.9 cents of gross profit on the incremental dollar sequentially, on an 11.1% book. Stock that rose in value because memory rose can fall in value for the same reason, with no change in units.
What the last two turns looked like
The build has never been what hurt. Avnet's sales peaked at $6,750.1M in , when inventory was 69% of quarterly sales. Inventory kept rising for five more quarters while sales fell, and the ratio reached 102% in . Operating income went from $313.6M in the quarter to $142.0M in the quarter, and only then did the cash come back in size. So the sequence is cash out, inventory overshoot, earnings down, cash back.
The harder precedent is 2001, when orders fell faster than return rights could absorb. EE Times reported then that Arrow wrote down $95M of inventory, which it called rare for franchise distributors, and that Avnet took a $327.5M special charge that included inventory it could not return (unverified, trade press, ). That book was a fraction of today's. We would not lean hard on one comparison from another era, and the franchise terms that protect distributors today are not disclosed in enough detail to say how much they would cover.
WT is where the arithmetic is least forgiving. A 1% write-down on NT$509.04B is NT$5.09B, or 49.4% of a record quarter's profit to the parent, and the December gross margin guide is 3.19% to 3.29%.
Why the resellers are a different story
CDW and Insight carry inventory of 15.1% and 10.9% of quarterly sales, both below their 2022 peaks of 17.7% and 14.5%. After those peaks CDW took inventory from $1,055M to $671M, and Insight from $383M to $145M, without a crisis at either. Their cushion is income that does not need a warehouse. Netted revenue was 38.2% of CDW's gross profit (unverified, third-party transcript), and services were 61.6% of Insight's.
Receivables are the line that looks alarming and mostly is not. They are 115% of quarterly sales at CDW against about 50% in the early 2010s, and 302% at Insight against about 85%, because netted deals carry the balance gross while the sale is booked net. That is a gross-up. It does make cash swing, and Insight's accrued liabilities rose $217.1M in a quarter with $128.4M of operating cash flow. Their real question is earnings quality: Insight's adjusted EPS was $3.59 against $0.74 of GAAP, and add-backs were 48.2% of CDW's non-GAAP operating income gain.
Super Micro does not fit either group. Its ratios look fine, and they are not where the risk is. One customer was 44% of September revenue, against 28.1% for the largest customer in fiscal 2026. Non-cancelable purchase commitments were $34.2B at June against $1.6B a year earlier, and the 10-K says some customers can cancel or defer on short notice. Operating cash flow was −$6.62B in March, in a strong market.
Everyone hedged on December
| Company | Guide | Stated reason or context |
|---|---|---|
| CDW | Gross profit up mid-single digits, against +10.3% in September | Calls pull-forward immaterial; the guide implies fewer large projects |
| Insight | Implied gross profit −0.8% to +4.7%, against +20.5% | Memory pricing, supply and macro; line guides sum to more than the total |
| Arrow | Sales $11.55B to $12.15B; Components +5.1% at the midpoint | Little memory price assumed, so units have to carry the step |
| Avnet | Sales $11.1B to $11.4B, about +12% after the extra week | Assumes similar supply, while late backlog grows |
| Ingram Micro | Sales $17.5B to $17.9B, about +10.5% Y/Y after its extra week | Thinner gross margin guided, on large AI deals |
| WT Micro | Revenue NT$730B to NT$770B, +26% at the midpoint | Almost entirely data center; other segments fading after pull-ins |
| Super Micro | Sales $19.5B to $21.5B; the full-year midpoint implies a smaller second half | Customer site readiness and component availability |
The resellers are guiding to slower gross profit, or building in a cushion, while the distributors and Super Micro talk about supply. Our read is a channel where parts are short and some end buyers may have bought early. If that is right, resellers would see a slowdown first, and the stock held further up the chain would be at risk after that. No company said this, and it is the piece of the page we hold most loosely.
So does RAM or a graphics card get cheaper soon?
Not soon. The prices that set what you pay are still going up, only more slowly. A DDR5 16Gb chip went from US$8.70 in to US$48.00 in on the TrendForce contract prices in Team Group's deck, up 451.7%. TrendForce's release projects conventional DRAM contract prices up another 10–15% in the fourth quarter and says the pace is moderating. It also says PC makers are still buying because they expect supply to stay tight in 2027, and that PC DRAM supply could decline next year as capacity shifts to servers (unverified, a single research firm's forecast).
Graphics cards follow, because their memory comes from the same makers. A German price tracker cited by TechTimes has the RTX 50 series up 66% on the year, and the same article says no major memory maker has projected relief before mid-2027 at the earliest (unverified, secondary report of a tracker; retail prices vary widely by region and seller). Those are street prices, and they move faster and further than contracts in both directions.
The six prints say the same thing from the other side. Insight's device revenue rose more than 30% on units up low single digits (unverified, third-party transcript), so buyers are already paying more for about the same number of machines. Press coverage of WT's call has consumer, PC and phone revenue falling in the fourth quarter after customers bought early (unverified, translated press). Weak consumer demand has not brought prices down, because AI servers take the supply that consumers give up.
So the thing that would make your RAM cheaper is AI orders slowing. That is also the thing that would leave these distributors holding memory bought at a high price. The relief consumers are waiting for and the turn this page worries about are one event.
Connecting the dots
Only June's 99.8% has been higher.
Stock for a quarter not yet reported.
Filing does not split out the ECS agreements.
A record in dollars, not in ratio.
Avnet and Arrow keep adding stock through December, because orders run past the guide and each has raised funding ahead of need.
- Evidence
- Arrow's backlog runs to mid-2027 by its chief executive's account (unverified, third-party transcript). Avnet added $1,000,000,000 of revolver commitments on . Last cycle inventory peaked five quarters after sales.
- Wrong if
- Their combined inventory is below $15.48B at the December quarter end.
- Worth
- Each 1% of that inventory is about $154.8M.
A DDR5 16Gb chip is still at or above US$48.00 on the TrendForce contract price at the end of .
- Evidence
- TrendForce projects a 10–15% rise in the fourth quarter and expects tight supply in 2027 (unverified, a single research firm's forecast). Team Group's deck says makers are giving capacity to HBM and high-end DDR5.
- Wrong if
- That contract price is below US$48.00 at the end of .
- Worth
- The first date a consumer could see relief start at the contract level.
The strongest case against this read
Here is the other side at full strength, and it is not weak. Arrow's adjusted leverage is 1.9x and Avnet's gross leverage is 3.1x, both falling because earnings are outrunning debt (unverified, third-party transcripts). Arrow's total debt of $2,752.0M is below the $3,129.5M of a year ago, on trailing non-GAAP EPS of $21.28 against $11.02 for 2025. Suppliers extended terms because they want these parts sold, and customers are prepaying at Super Micro and were at WT, which is not how a glut behaves. On that reading we have spent a whole page on a working capital line that reverses by itself, as it did in 2024.
Most of that may be right about the business. We still land where we do on the narrow question, because none of it changes which way inventory moves next quarter. If Arrow and Avnet ship above their guides, our call falls, and that is what the $15.48B line is for.
What we'd watch
By . WT's October revenue against NT$243.33B a month for the low end of the guide.
By . WT's Q3 statements: contract liabilities against NT$35.95B, and where the US$500M convertible sits.
The 10-Qs. Avnet's sold receivables against $1.98B in June, and Super Micro's purchase commitments against $34.2B.
Payable days and payables ÷ inventory. A fall while inventory holds is the first sign suppliers are pulling terms, and the reading that would change our mind on how serious this is.
Memory pricing. Components gross margin below 11.1% at Arrow would be a fourth straight fall, the first with little memory price assumed.
End of January. TrendForce's DDR5 16Gb contract price against US$48.00.
December quarter-end inventory. Avnet plus Arrow at or above $15.48B, reported around early February (unverified, dates not confirmed by the companies).
Bottom line
This is not yet a dire situation, and calling it one would misread how distributors have always financed an upcycle. The burn and the build are the norm. What has changed is the cushion underneath them: gross margin at or near the lows, supplier credit at levels the series has not shown, and a third of the growth at Avnet that is memory price. Those do not hurt on the way up.
They decide how much the turn costs. Our call is that the turn is not December: Avnet and Arrow together end the quarter with at least $15.48B of inventory. Wrong if they end below it.