🌐 The Masking Filter: The Disinflation That Wasn't, Memory's Contract Gambit & Germany's Great Sorting
The Ruck Filter #032 • August 17, 2026
Read time: 8 minutes
Welcome back to The Ruck Filter.
Wednesday’s inflation print was received as good news, and on the surface it was. Consumer prices rose just 0.1% in July, putting the annual rate at 3.4%. Core came in at 2.5%. The S&P 500 hit a record on Thursday, September rate-hike odds fell from roughly 55% to about 34%, and the VIX dropped below 14.4, a new low for 2026.
I spent the week looking at where that disinflation actually came from, and I think most people are reading it backwards.
Energy fell 1.5% on the month, with gasoline down 2.9%. That single line did most of the work. Meanwhile the largest goods-price impulse in years is building in a part of the economy the CPI has barely started to register and on Thursday the industry at the centre of it announced a plan to make those prices permanent.
Then Friday arrived and undid the mood. Retail sales fell 0.6%, the sharpest drop in more than a year, against expectations of a small gain. Consumer sentiment came in at 51.0.
Three sections: why this disinflation is borrowed, what memory just tried to do to its own cycle, and how Germany spent the week sorting its winners from its losers inside the same sectors.
1. Signal vs. Noise: A Disinflation Built on Oil, and Memory’s Contract Gambit 💾
The Noise: Inflation is coming down, the Fed can stand still, and the goods-price scare is over.
The Alpha: Strip out energy and the picture changes. The disinflation is borrowed from a barrel of oil that has already turned. Underneath it, memory prices are running at a pace that is showing up in laptops and storage drives rather than in the CPI’s headline. And this week the memory industry made an extraordinary move: it tried to contract its way out of its own cycle. That doesn’t abolish the risk. It converts it into something less familiar and, I’d argue, less well priced.
The Filter: Start with the composition. July CPI rose 0.1% on the month for a 3.4% annual rate, matching consensus, with core at 0.2% and 2.5%. Energy fell 1.5% and gasoline 2.9%. Shelter rose only 0.1% and still accounted for roughly two-thirds of the entire monthly increase, which tells you how little else was contributing. Producer prices came in mixed. Rate markets took the hint: the two-year yield eased from about 4.2% to 4.15%, and September hike odds roughly halved.
Now the part that isn’t in that number yet. Memory prices have been running hot for months. Counterpoint’s data earlier this year showed the DRAM market posting 30% quarter-on-quarter growth for two consecutive periods on price alone. Gartner has forecast PC prices rising 17% this year. Solid-state drives cost two to three times what they did in December. That is the AI cost boomerang I described back in Issue #025, and it has not gone away. It has simply been outrun by falling energy in the headline calculation.
Which brings me to Thursday. SanDisk held an investor day and, instead of guiding a quarter, published a model for fiscal 2028 through 2030: revenue growth in the mid-to-high teens annually, an adjusted gross margin around 80%, and a policy of returning all excess cash after investment. The number that matters most is buried in the mechanics. SanDisk disclosed multi-year contracts with eight companies, including three US hyperscalers, explicitly framed as an effort to dampen the price volatility that has always defined NAND.
The market reaction was violent. SanDisk rose about 13.6% on the day. Micron climbed toward $966, approaching a $1,000 handle. SK Hynix’s ADR gained more than 7%, Seagate about 5%, and Korea’s Kospi jumped 4% overnight, leaving it 23% above its July 30 low. All of this after a July in which the memory complex had drawn down somewhere in the region of 30% to 40% on oversupply fears.
Here’s what I think is genuinely new, and what I haven’t seen framed anywhere. What SanDisk did is install a meter on the supply side. In Issue #030 I argued the market rewards capital spending when there’s a contracted, metered revenue line attached to it. Memory makers have now applied that logic to themselves: lock the demand into multi-year agreements, and a commodity becomes a utility. The market paid for it immediately.
But a contract doesn’t destroy risk. It relocates it. An 80% gross margin held through fiscal 2030 is a target, not an achievement, and it depends entirely on three hyperscalers honouring multi-year commitments. Those hyperscalers are the same firms that have committed $720 to $745 billion of capital spending in 2026 against a payback nobody has yet demonstrated. So the memory cycle hasn’t been abolished. It has been converted from price risk into counterparty risk — and counterparty risk is much harder to see coming, because it doesn’t show up gradually in spot prices. It shows up all at once, in a renegotiation.
One detail keeps me honest about which way this cuts. Part of what triggered July’s selloff was Micron’s CEO filing a Rule 10b5-1 plan to sell stock near the highs. In Issue #027 I wrote about what it means when the owners of a shortage start selling it. That signal and this week’s contract announcements point in opposite directions, and I don’t think the argument is settled.
A measured approach:
Read the CPI by component, not by headline, for the rest of this year. The disinflation on offer is energy-shaped and reversible. Oil rose again into Friday. What to watch: the goods-versus-services split in the next two prints, and specifically whether household electronics start contributing.
Treat long-term memory contracts as a credit question, not a pricing question. The right due diligence on SanDisk’s 2030 model is not a supply-demand forecast. It’s an assessment of whether three hyperscalers will still want this capacity at these prices in 2029. What to watch: contract disclosure detail - take-or-pay terms and duration - in the next round of memory earnings.
Nebius is the clean read on whether the contracted model holds. Its quarterly revenue rose 454% to $582.3 million with remaining performance obligations of $37.49 billion against first-half capital spending of $8.13 billion. That ratio of booked obligations to spending is the single most useful number in the AI infrastructure complex right now.
2. The Consumer Who Isn’t There 📉
The Noise: Earnings are strong, 87% of companies are beating, the market is at records. The economy is fine.
The Alpha: Friday’s two data points describe a consumer in worse shape than at almost any point in the past two decades, while the index sits at an all-time high and volatility trades at its lowest level of the year. I flagged this gap in Issue #031. It got wider this week, and the complacency is now measurable.
The Filter: Retail sales fell 0.6% in July. Consensus was somewhere between a 0.1% and 0.2% gain. That’s the largest monthly decline in over a year. Some of it is mechanical - lower petrol prices reduce the dollar value of fuel sales, and car sales fell - but the miss against expectations was wide by any reading.
The sentiment number is the one I keep coming back to. The University of Michigan’s preliminary August index came in at 51.0, down from 55.2 in July, against forecasts of 55. Its long-term average sits around 84. To be clear about what that means: American consumers are reporting a level of economic pessimism that historically shows up in recessions, not in the middle of a record-setting equity rally.
Set that against the corporate picture. With 455 of 500 S&P 500 companies reported, 87% beat earnings estimates and 68% beat on revenue. The S&P is up about 14% for the year and the Nasdaq about 15%. The VIX fell below 14.4 this week, a 2026 low, which tells you how little downside hedging is being bought.
I want to name the mechanism rather than just the contrast, because the contrast alone is a talking point. Corporate earnings this quarter have been carried by the AI complex and by companies serving enterprise capital budgets. Consumer sentiment measures household finances. Those are two different customer bases, and for eighteen months they have been diverging. That works right up until it doesn’t, because eventually enterprise capital budgets are funded by revenue from someone, and a large share of that someone is a household.
The near-term test is unusually well timed. Walmart and Target report next week, before Nvidia on the 26th. Those two will tell you more about whether the sentiment reading is noise or signal than any macro release. Then Jackson Hole runs from 27 to 29 August, where a new Fed chair with three dissenting hawks on his committee has to explain a policy stance to a market that has stopped pricing hikes.
A measured approach:
A VIX under 15 with sentiment at 51 is an unusual combination and worth respecting. Cheap volatility is not a prediction, but it does mean protection costs less now than it has all year. For portfolios that have ridden a 14% index gain, this is the least expensive moment in 2026 to think about the downside.
Watch the retailers, not the economists. What to watch: Walmart and Target guidance language on trade-down behaviour and unit volumes rather than headline sales. Unit volumes strip out the price effects that make revenue look healthier than demand.
The divergence itself is the position. Businesses selling into enterprise capital budgets and businesses selling into household budgets are in visibly different cycles. Owning both without distinguishing between them is the error this data is warning about.
3. Germany’s Great Sorting 🇩🇪
The Noise: German defence is one trade and German semiconductors are another. Pick the sector and ride it.
The Alpha: This week Germany produced two textbook cases of the same lesson: inside a single sector, in a single week, the winners and losers moved in opposite directions for entirely legible reasons. In defence a naval shipbuilder tripled its guidance while its own sister company was cutting. In semiconductors, companies shipping real volume rallied alongside one shipping almost none. Sector exposure is now the lazy version of the trade.
The Filter: Start with defence, because the contrast is almost too neat. TKMS, the naval shipbuilder in the MDAX, reported on Wednesday and raised its full-year revenue growth guidance from 2–5% to 10–12%. Nine-month revenue rose 19% to €1.89 billion, adjusted EBIT rose 13% to €110 million, and in the submarines segment adjusted EBIT quadrupled to €46 million as higher-margin new-build projects ramped and legacy burdens rolled off. Atlas Elektronik grew revenue 28%. The order backlog stands at roughly €20.1 billion. Adjusted EBIT margin is now guided up to 6.5%, with the medium-term target of about 10% annual growth and more than 7% margin confirmed. The stock rose about 8.5% on Xetra, and reporting elsewhere put the move as high as 11.75% to €98.90. Bernstein upgraded it.
There’s a detail in the German coverage that I found more interesting than the numbers. The Iran conflict is generating new demand from the Persian Gulf for naval systems. A maritime chokepoint war creates demand for maritime capability. Obvious in hindsight, and yet.
Now the mirror. Rheinmetall, part of the same national defence build-out, had to cut its 2026 revenue guidance after the F126 frigate programme was cancelled. I flagged that cancellation in Issue #026 as evidence that procurement politics cuts both ways in a maturing defence trade. This week it showed up in guidance. Two German defence names, same fiscal tailwind, opposite direction, separated by product mix and by which government committee happened to decide what.
The semiconductor chain tells the same story with a sharper edge. Aixtron is up roughly 142% year to date, including about 45% in six trading days earlier this month. Süss MicroTec is up around 114%, Siltronic around 79%. Elmos Semiconductor reported first-half revenue up 15.4% to €314.5 million at a 24.1% EBIT margin. Infineon was among the DAX leaders early in the week after a Goldman Sachs upgrade. That is a chain with real volume behind it, and it’s the continuation of the wafer story I wrote about in Issue #030.
Then there’s LPKF, and I include it deliberately as the counterweight. On the screen it looks like the best German stock of 2026, up somewhere around 175% year to date. The fundamentals underneath are the opposite. First-half revenue fell 38.3% to €36.5 million from €59.2 million, and EBIT came in at minus €14.1 million. The solar segment is suffering because customers are holding back investment while the shift to perovskite cells hasn’t arrived. The bull case rests on LIDE, its laser deep-etching process for advanced packaging, where management now puts the addressable market at about €1.7 billion by 2030. In the second quarter it shipped two systems: one to a specialty glass manufacturer and one to Penn State University. The company is mid-restructuring under a programme called North Star, absorbing costs of 3–4% of revenue this year, targeting a double-digit EBIT margin by 2028. The CFO bought 1,500 shares at €14.00 in late July, which I read as a genuine signal but a small one.
I’m not calling LPKF a short and I’m not qualified to judge LIDE’s engineering. What I’d say plainly is this: a company with two shipped systems in a quarter and a €1.7 billion market claim for 2030 is a venture bet wearing a listed company’s clothing. It belongs in a completely different risk bucket from Siltronic or Elmos, and the year-to-date table on your screen will not tell you that.
A measured approach:
Sort German industrials by product mix and end customer, not by sector label. That was the argument in Issue #029 with trucks against cars. This week ran it twice more, in defence and in semiconductors. It is becoming the single most useful screen for DACH allocation this year.
In defence, the naval and maritime segment has a demand driver the land systems don’t. What to watch: whether TKMS delivers the medium-term target upgrade it has signalled for next quarter, and whether Persian Gulf interest converts into signed orders rather than inquiries.
In the chip chain, separate volume from narrative. Siltronic, Elmos and Aixtron are shipping into real demand. LPKF is selling a roadmap. Both can rise in a thematic bid. Only one survives a thematic unwind. What to watch: LPKF’s Q3 report in October for whether LIDE produces a genuine volume order.
Outro: What Gets Masked
Three maskings this week, and I think they rhyme. Falling oil masked a goods-price impulse still building underneath. Long-term contracts masked a cycle by converting it into counterparty credit. And a record index masked a consumer reporting recession-level pessimism, while a year-to-date performance table masked the difference between a company shipping wafers and one shipping two machines.
None of these are deceptions. They’re just aggregates doing what aggregates do, which is average away the thing you actually needed to know. The work - the only work, really - is disaggregation. Read the CPI by component. Read the contract by counterparty. Read the sector by product mix. Read the index by who its customers are.
That’s slower than reading the headline. It’s also the entire reason this newsletter exists.
The Takeaway: When the headline number and the composition underneath tell different stories, which one is in your portfolio?
Daniel Ruck
Editor, The Ruck Filter
Filter Sources this week
US Bureau of Labor Statistics | July CPI report, August 12, 2026 (via Stock Rover, CNBC)
Edward Jones | daily market recap, CME FedWatch pricing and retail sales, August 13–14, 2026
Charles Schwab | market update, VIX levels and earnings scorecard via Bloomberg, August 14, 2026
Yahoo Finance & Trading Economics | Friday index closes and University of Michigan sentiment, August 14, 2026
SanDisk | investor day long-term financial model, August 13, 2026 (via moomoo, AOL, TradingKey)
24/7 Wall St. | memory and storage sector coverage, Kospi and ETF flows, August 13, 2026
Yahoo Finance | Nebius Q2 results and memory rally drivers, August 12, 2026
CNBC | Counterpoint DRAM data and Gartner PC price forecast, April 2026
TKMS | nine-month results and guidance raise, August 12, 2026 (via ARIVA, Platow, finanzen.net)
Börse Express | Bernstein upgrade on TKMS and Rheinmetall F126 comparison, August 14, 2026
Finanztrends & Börse Global | German small-cap performance data, LPKF half-year figures and LIDE detail, August 10–14, 2026
Disclaimer: The Ruck Filter is for informational purposes only and does not constitute financial, investment, or tax advice. The information provided is based on data available at the time of writing and is subject to change. Investing in financial markets involves risks, including the potential loss of principal. Every reader is solely responsible for their own trading and investment decisions. Please conduct your own due diligence or consult with a licensed professional before making any financial commitments. Companies with potential conflicts of interest are excluded from coverage as a matter of editorial policy.