AI’s $204 Billion “Security” Tax
· InvestorPlaceListen to the audio version of this article (generated by AI).
PCE inflation comes in cooler than expected… why tonight’s Micron print is important… the AI “security tax” that could be a tailwind… Eric Fry’s safer way to play the boom
It was a busy morning on the macroeconomic calendar…
We’ll start with the most important piece of data – the August Personal Consumption Expenditures (PCE) price index, which is the inflation gauge the Federal Reserve watches above all others. It came in cooler than expected.
Headline prices rose 3.4% from a year ago, below the 3.7% forecast. Core PCE – which strips out food and energy – rose 3%, below its estimate of 3.3%. Core has now held at 3% for three straight months.
While that steady core reading is good news, the bad news is that the headline reading doesn’t capture September’s surge in diesel prices. So, next month’s headline print is likely to look much hotter. Still, this is welcome news for investors who have been worrying about rocketing treasury yields and the potential for another Fed hike in October.
Turning to the economy, second-quarter GDP was revised sharply higher, coming in at a 2.2% annual growth rate, up from the earlier 1.5% reading. The change was due to firmer consumer and government spending.
Speaking of the consumer, inflation-adjusted spending jumped 0.6% in August, the biggest monthly gain since early 2025. This came even as incomes barely budged and the savings rate slipped to its lowest since 2022. Another win for the “resilient” U.S. consumer who refuses to close their wallet.
Finally, this morning’s jobs data came in reasonably strong. Private payrolls rose 90,000 in September, according to payroll processor ADP. That was better than the 68,000 expected and the strongest reading in three months. Of course, Friday’s official BLS employment report is the big one to watch.
Returning to this morning’s cooler inflation number, it takes some of the urgency off an October hike. According to the CME Group’s FedWatch Tool, traders have lowered their bets for a quarter-point hike next month from 51% yesterday to about 35%. This optimism is helping push stocks higher while keeping treasury yields flat as I write approaching lunchtime.
Bottom line: the macro picture bought the bulls a little breathing room this morning. But don’t get too comfortable – in fact, by the time you read this, it’s likely a new number will have dropped that has market-rattling potential…
How did Micron do?
By the time this hits your inbox, Micron Technology (MU) will likely have reported its latest quarterly earnings.
If you’ve read the Digest over the past month, you know why this is important. Twice now – first with Micron, then with Sandisk (SNDK) – we’ve laid out the same puzzle: a memory maker printing record, AI-fueled profits, yet priced by Wall Street as if the floor is about to give way.
The market can’t shake its old fear that memory is a boom-and-bust business destined to crash. I’ve challenged that assumption, but I won’t re-litigate the whole case today.
Even if you don’t own MU, its performance tonight serves as a helpful diagnostic for the entire AI trade. After all, strip the AI build-out down to its studs, and it’s three things – compute, memory, and power. Micron sits squarely on the memory leg. So, if demand is really holding up the way the bulls insist, it should show up here, tonight, in hard numbers.
That’s important in this type of skittish market – sentiment swings by the hour, but earnings are what will help instill confidence for a more durable climb.
So, check the after-hours numbers, and we’ll dive into the results in a future Digest.
The AI “security scare” everyone’s misreading
Now, let’s turn to the story the financial press has been hammering for weeks: AI security.
The headlines have been relentless – breaches, rogue agents, frontier labs pausing training to patch holes. The knee-jerk read is that this is bad news for the whole AI trade. More danger means more caution, more caution means slower spending, and slower spending means the AI infrastructure names take a hit.
Our technology expert Luke Lango, editor of Innovation Investor, thinks that read has it backward. All this new security might be a tailwind hiding in plain sight.
Here’s Luke from Monday’s Daily Notes:
OpenAI’s estimate that enhanced monitoring consumes roughly 20% of the inference compute of monitored workloads gives us a concrete way to think about a new, largely unmodeled layer of AI infrastructure demand…
If safety monitoring becomes standard practice across frontier labs and enterprise deployments, every agentic task effectively carries a supervisory compute overhead on top of its productive compute – overhead that grows alongside usage rather than replacing it.
If keeping AI safe eats 20% more compute on top of the work itself, then every dollar spent on safety is another dollar of demand for exactly the compute, memory, and power we just nodded to when discussing Micron.
Overall, Luke frames the safety issue as being a compute tax, not a compute cut.
And Luke isn’t alone in his take. Independent researchers have started formally measuring this “control tax” – the compute cost of monitoring AI – and Gartner now projects the AI-security market alone will balloon from about $49 billion this year to $204 billion by 2030.
There’s a virtuous circle underneath it, too. Back to Luke to explain:
So, better controls enable broader deployment. Broader deployment generates more usage. And more usage requires both productive and supervisory compute.
Overall, the scare that’s supposedly a headwind for AI may quietly be building a whole new layer of demand.
To follow along with Luke and get his favorite ways to play this, click here to learn about joining him in Innovation Investor.
Eric Fry’s backdoor way to play the boom
Now, whether Micron crushes it tonight or stumbles… and whether AI security turns into a real tailwind or just noise… you’re still betting on the companies building the AI machine. And, of course, that carries elevated risk.
So, our global macro expert Eric Fry, editor of Fry’s Investment Report, has been pointing at a different – and potentially less risky – way to play AI.
His model is the dot-com fiber boom, when telecom firms sank close to $1 trillion laying 80 million miles of cable. When the boom dried up, most of the builders went bust. But a second group of companies got rich off it – the ones Eric calls the “appliers”:
They never touch a shovel or a spool of cable. They simply wait for the new technology to become ubiquitous and cheap, then capitalize on it by finding ways to apply the new technology to their own business.
His example was Humana (HUM) – a health insurer that digitized its paper claims onto all that new fiber and went on to soar more than 5,000% over time. It didn’t build the internet. It applied it.
Eric sees the same setup forming in AI right now – and I want to share with you one of Eric’s recommendations for how to play it: Novo Nordisk A/S (NVO),
Eric’s lead analyst, Tom Yeung, dug into it in his recent Fry’s Investment Report Weekly Update. As he explained, Novo is using AI to accelerate its drug development – and it’s funding that work with the rich cash flows from an already-dominant franchise…
Its new $299-a-month oral Wegovy weight-loss pill has been one of the fastest launches in the company’s history: 12 weeks to the first million prescriptions, then just four weeks to add the next million. The pill helps users shed about 16.6% of their body weight – ahead of the 12.4% achieved by Eli Lilly’s (LLY) oral candidate – with no weekly injections required.
That moat is what makes AI applications so powerful.
Cheap AI is becoming available to everyone, so “using AI” doesn’t necessarily buy a lasting edge. But what can’t be copied is what Novo aims AI at: a drug franchise walled off by at least eight patents and decades of know-how. No competitor can suddenly churn out Wegovy pills tomorrow, no matter how good their AI is.
Here’s how Eric put it:
Novo’s AI edge isn’t exclusive, but it doesn’t need to be…
It’s about recognizing that a company applying cheap, borrowed AI compute to an already-profitable, already-growing business captures real value, while the companies building and financing the underlying AI infrastructure absorb the risk.
Or, as Tom sums it up:
Novo already has the non-commoditized products that competitors can never legally replicate. And so, cheap AI compute is just icing… on a cake no one else can bake.
Coming full circle
Step back from the day, and it’s hard not to feel a little optimistic. Inflation came in cooler than feared… growth was revised higher… the consumer refuses to stop shopping… and even AI’s scariest headline may become a new source of growth.
After weeks of bracing for the next blow, the bulls finally caught a few breaks in a single morning.
But as I said earlier, the next test is already here in Micron. So, go see how the numbers came in – it’ll likely tell us about tomorrow’s market.
We’ll report back.
Have a good evening,
Jeff Remsburg
(Disclosure: I own MU)