Here’s a preview of what we’ll cover this week: 

Macro: Fed Backs The US Economy; Inflation Is Still the Problem

Markets: Markets Finish Higher as Leadership Flips; How Is The Earnings Season?; Microsoft: Please Don’t Associate Me With OpenAI; OpenAI: it keeps getting worse; Meta: Strong Core, Bigger Bill; Apple: The Dip Was Due; Where Are We In The Rotation?

Lumida Curations: Sam Altman: Intelligence Is Becoming a Commodity; Dan Ives: Cybersecurity Budgets Are About to Double; Akshay Nathan: AI Is Blurring the Lines Between Every Job

This week, I did a podcast with Michael Parekh, talking about markets, memory, and latest earnings. Michael was the lead internet analyst at Goldman Sachs during the dot-com era, and essentially a peer of Mary Meeker and Henry Blodgett, who were at Morgan Stanley and Merrill Lynch.

Extremely thoughtful individual. I always love having him on

Here’s what we discuss:

The podcast comes out this Sunday on 11 am ET on our Youtube Channel — Lumida Wealth. Watch the podcast here.

Situational Awareness led headlines this week, as they sold their portfolio to Citadel after incurring massive losses. 

Could anyone have foreseen it?

Not to brag, but I flag the exact risk that led to Situational Awareness’s downfall back on June 11th, when SMH was trading around 650 (2% below its top, and 15% higher than its current level).

This week, I did a FSD stream, titled “Ladies, Liquor, and Leopold”, discussing what happened to Situational Awareness, and where it went wrong. Watch it here. 

Leopold’s portfolio appears to have peaked in early June, around the time Forbes published this headline:

“24-Year-Old AI Wunderkind Is Betting Big on Bitcoin Miners.”

Roughly 2 months later, we are here, with WSJ publishing a very different headline:

“Citadel Buys Situational Awareness’s Stock Portfolio.”

The chart below helps explain what happened in between.

The above screenshot comes from the Lumida Invest app. 

You can see how Leopold’s portfolio was essentially a massively crowded bet on a single, high-beta concept. 

Notice how it is overexposed to animal spirits, and momentum, while carrying negative exposure to all quality factors. 

The simple factor chart was the reason I could identify Leopold’s one-sided positioning risk one month before the market charged him for it.

It’s interesting how factor exposures can get you insights that transcripts and technicals can’t. 

(Soon, our app users will be able to see the factor exposures for their portfolios: which factors they actually own, where their risk is concentrated and whether ten different positions are secretly one trade.)

Now, I have a theory about what and how it went all wrong for Leopold.

Gather round…

1. He started on the wrong side of the market

We can see Leopold’s positions as of March 30 through 13F disclosures.

Leopold had puts on Nvidia, AMD, Oracle, and Broadcom, and they are all of significant sizes. 

These positions are right around the market’s bottom. (See chart below)

He was positioned the wrong way.

2) Then, came April, and all of these names rallied hard. So, his puts became worthless. 

3) And, now, he needs to cover his positions, so he starts covering at a loss. Names like AMD were up 50 to 70% as he is covering on the way up. 

He keeps his software shorts.

4) Now, he’s fully long and the semis are up significantly, but he got left behind. And, he needs to make it all back. 

5) He levers up 4x and goes giga long semis. During the same time, he goes short on software stocks to fund his positions, taking inverse positions in names, like Salesforce, Adobe, and Intuit — all of which are reasonably valued and fundamentally sound companies. (Remind you, this is our theory, looking back the tape in hindsight)

6) His moves now create a meteoric rise in semis.

What we saw in semis was an extremely rare < 1% move in market history. 

The category was overbought and then did another melt-up as if it was just warming up. 

That second leg was all Leopold. 

7) Then, June rolls around. The SMH index started showing first signs of fatigue.

Remember, we flagged the above ETF flows chart on June 21st. SMH peaked that week. 

Everyone got in memory… including the humble nation of South Korea to which Leopold should be granted dual citizen status. 

No one else was left to buy. 

8) Then, came July, and we saw the mighty rotation. His long leg, semis, started to cool off. Micron’s report became the ‘sell the news’ signal. 

On the other end, his shorts were exactly where the funds flowed to. Software names started ripping higher, and those shorts blew up in his face.

9) Leopold likely bought the dip on those strong semis numbers…software continuees to rally, then he became history.

Leopold says the blowup happened after the spread between his longs and shorts reversed.

That is true.

But, the move was not unique to his portfolio.

The momentum factor and the animal-spirits factor both crashed after June, as we noted in our June 28th newsletter, “The Aftermath of the SpaceX IPO.”

Leopold’s beta was in the wrong places. 

During July’s rotation, earnings yield factors (think financials, banks and small cap value) moved higher. 

Biotech (a form of animal spirits) moved higher, too, but his beta was on AI which did not. 

These factor moves happened across the entire market. 

No one was targeting or ‘hunting’ Leopold’s positions… at least not until the last two weeks when markets sniffed out forced deleveraging. 

Quants were also selling momentum. 

Here’s our proof.  

We have this model portfolio in the Lumida Invest app called ‘Extreme Momentum (QQQ)’.

It recorded a 11% dip in July, which if you are 4x levered, can erode roughly 50% of our portfolio, which is what we believe happened with Situational Awareness. 

‘Extreme Momentum (QQQ)’ is one of the many strategies we have added to the Lumida app. 

The only thing the strategy will ever buy is momentum. It’s been 70% cash for a while now. It’s struggling to find momentum stocks in the QQQ. 

Have a look at its P&L curve below. The chart matches the peak in momentum factor, but the drawdown was negligible. 

Also, take a look at the strategy’s sales history.

Notice how the strategy was exiting semis exactly when Leopold should have been. 

Could Lumida Invest app have saved SA? My guess is yes!

You should download the Lumida Invest app, and check out our ‘Strategies’ page. Then find ‘Extreme Momentum (QQQ)’ to replicate what we show above.

We have strategies to locate momentum, mean reversion, quality, animal spirits and more. 

You can also see the strategy’s performance across multiple timeframes, and relative to the index. The trade history can help you see what the strategy is buying next. Note: These exclude transaction costs and are algorithmic backtests. I use these as a tool for idea generation.

It’s a glimpse into the financial advisor for the future!

Luke sees our vision.

Tell me: how does the 2/20 model survive the age of AI?

The Changing of the Guard

There is another part of this story that deserves attention.

When Long-Term Capital Management collapsed during the dot-com bubble, the New York Federal Reserve summoned Goldman Sachs, JPMorgan and Morgan Stanley to help organize a private bailout.

Leopold’s fund was never systemic. But look at the firms reportedly involved this time:

Citadel. Jane Street. Millennium.

No traditional sell-side investment bank.

Not BlackRock.

Not Blackstone.

Three quantitative investment firms that have already transformed their operations with AI.

Jane Street already generates more profit than Goldman Sachs with a fraction of the headcount.

Both Jane Street and Citadel are massive consumers of GPU compute.

This is more than a hedge fund changing hands.

It is another sign of the changing guard.

Traditional financial institutions were built around people, relationships and balance sheets.

The next generation is being built around algorithms, proprietary capital and compute.

The AI-native firms are no longer sitting at the edge of the financial system.

They are becoming the system.

The Federal Reserve held rates at 3.50%–3.75%.

The vote was 9–3, with Beth Hammack (Cleveland Fed chair), Neel Kashkari (Minneapolis) and Lorie Logan (Dallas) preferring a 25 bps hike.

Chairman Kevin Warsh’s conference had some insightful nuggets on the economy.

Start with broad economic view: “The economy is showing impressive resilience.”

He remarked: Economic growth remains solid. Job creation is keeping pace with the workforce. And unemployment has barely moved.

This week’s GDP and unemployment data backs him up.

Q2’s final sales to private domestic purchasers, the cleanest measure of underlying demand, rose 3.9%, the strongest increase since early 2023.

Consumer spending also accelerated to 3.2%, up from just 0.5% in Q1.

Nonresidential fixed investment also increased 8.4%.

The weakness in Q2’s headline GDP (1.5% vs 2.1% in Q1) largely came from trade. 

Imports surged ahead of tariffs, subtracting roughly 1.5 percentage points from growth.

Strip that out, and domestic demand was strong.

The consumer is not rolling over either.

Real consumer spending rose another 0.4% MoM in June, reaching a new record high.

The gains were broad-based.

Hotels, recreation, apparel, vehicles and other discretionary categories all moved higher.

The labor market is also as solid as ever.

Initial unemployment claims rose modestly to 197,000, but the four-week average fell to its lowest level since January 2024.

Continuing claims declined for a third consecutive week.

The economic engine is working.

But!

When everything driving demand is better than before, you have one problem.

And, the Fed knows it. 

Warsh’s comments on economic resilience were overshadowed by high and sticky inflation.

He tried the same mantra as his first conference, and focused on controlling yields by sounding serious about inflation.

“There is no soft inflation target, there is no soft implicit target… There is only a target, and it is 2 percent.”

However, unlike the last time, bond market didn’t buy Warsh’s comments, and the call to ‘hold’ over-ruled his efforts on Inflation accountability.

US 10Y rose to 4.74% on Friday, the highest level since Jan 25. Rate hike odds also increased to 67% from 55% a week earlier. 

Unfortunately, the latest data gave him little reason to relax.

In June, headline PCE inflation stood at 3.7% YoY, while core inflation remained at 3.3%.

Goods inflation eased slightly to 3.7%, helped by a 9.6% monthly decline in gasoline prices.

But tariff-related pressures have not fully passed through.

And the AI buildout is pushing up demand for chips, memory, data-center equipment and power infrastructure.

So, where does this leave us on rates?

Our base case is no hike (or cut) in 2026. 

AI is creating two opposing forces.

In the short run, the buildout is inflationary.

It increases demand for chips, memory, power, construction and specialized labor.

But over time, AI is a deflationary force.

It allows businesses to produce more with fewer people.

It lowers the cost of software, research, customer service, administration and decision-making.

It improves productivity.

And productivity is ultimately how an economy grows without generating the same amount of inflation.

Our view is that the productivity effect will become more visible over the coming quarters.

The second variable is energy.

If relations between the U.S. and Iran continue to normalize, the geopolitical premium embedded in oil should decline.

Lower energy prices would flow through transportation, goods, manufacturing and headline inflation.

That would remove one of the biggest near-term reasons for the Fed to hike.

Markets ended the week higher. The S&P 500 gained ~1%, while the Nasdaq rose ~0.5%.

At the index level, the week looked calm.

Underneath the surface, it was anything but.

Software was the best-performing industry, with IGV up 7.5%.

Semiconductors were the worst, with SMH down 3.7%.

A few weeks back, this would be impossible to believe.

We do believe software is in a bull market and we recommend reading Porter Stansburry’s criticism of Leopold here.

Software’s move was helped by the strength in mean reversion factor.

IGV had been beaten down during semis trade (-20%), and it’s only natural that it receives the flows during this rotation.

Look at mean reversion strategy performance, and see how it has cranked higher, while the index has stayed flat. 

If you’d like to explore the mean reversion strategy, and see what it holds. Check out the Lumida invest app. It’s available on both appstore and playstore.

Soon, we will be launching automated portfolios, where AI automatically deploys and manages your funds according to the selected strategy.

What about Semis?

We saw high volume this past Wed in the SMH index just north of the 200 MA. The category may have bottomed.

That said, names in the index like AMD remain over-priced. It could very well take a few months to settle. And, mean reversion names are doing well.

The Leopold liquidation was likely the clearing event.

It pushed SOX volatility up to 75% - the only previous two time SOX volatility reached such levels was back in March 2020, and April 2025.

And, after both instances, the volatility spike marked the immediate bottom, and the index soared higher thereon.

The leveraged positions built around semiconductors, memory and AI hardware have now been largely washed out.

The Korean leveraged ETF boom has almost completely unwound, reaching its former averages.

Retail participation has collapsed too.

Net retail buying has fallen close to its lowest level since the pandemic.

That is usually when markets become interesting — When everyone who wanted to sell has already sold.

At some point, the absence of buyers stops being the problem.

The absence of sellers becomes the catalyst.

Individual names are also reaching valuations where fundamental support becomes reasonable.

SK Hynix now trades at roughly 4 times forward earnings.

Micron trades at about 6.1 times.

These are cyclical multiples being applied to companies sitting at the center of the AI memory buildout.

Memory remains a bottleneck. Short-term timing is hard, but the multi-year story is intact.

More than 700 companies have reported so far. And, headline earnings have been strong.

78% of the companies have beaten earnings estimates, while 74% have beaten on revenues.

Guidance has also been constructive: 14% of companies raised their outlook, while just 3% lowered it.

But, investors have not been handing out participation trophies.

The median stock opened only 0.2% higher after reporting, gave most of that back during the session and finished the full day up just 0.1%.

Four of the Mags reported this week.

And, let’s start with our favorite of the four. 

Microsoft. 

We’ve been talking about Microsoft is nearly every or every other newsletter for the past 2 months. The value was simply too compelling. The results validate our thesis.

Google’s earning move replicated Microsoft last week, so Microsoft returned the favor this week. 

Microsoft reported strong numbers, defended the CapEx and ripped up 15%.

It was MSFT’s first positive move on earnings after 3 consecutive quarters of an earnings day gap down.

Headline numbers were Microsoft-esque.

Revenue reached $90 billion, up 18%.

Operating income increased 18%.

Microsoft Cloud revenue increased 27% to $59.3 billion, while Azure accelerated to 43% growth.

Amy Hood (CFO): “Demand continues to exceed available supply.”

Azure is expected to grow 45% next quarter, with demand still ahead of supply.

Of course, the spending is enormous.

Microsoft spent $41 billion on capital expenditures this quarter and expects calendar-year CapEx of roughly $175 billion. 

But the composition matters.

Roughly two-thirds of Microsoft’s CapEx went towards shorter-lived assets, primarily CPUs and GPUs.

That gives the company more flexibility than a headline CapEx number suggests.

If demand slows, Microsoft does not have to abandon a half-built city.

It can simply buy fewer chips.

Hood explained:

“If the demand environment changes, you just slow down what is, in fact, the largest component.”

Land purchases and data-center construction can also be delayed or staggered.

Microsoft is spending aggressively.

It is not spending blindly.

The most interesting part of the call was not Azure growth.

It was how deliberately Microsoft separated its own economics from OpenAI, and other frontier labs.

The language in financial numbers was not subtle.

Microsoft’s release noted:

Management remarked all sequential RPO growth came from customers outside frontier-model companies. And, nearly 90% of Microsoft Cloud revenue now comes from customers outside the frontier labs.

Management repeated “excluding OpenAI” too many times for it to be accidental.

Microsoft was telling investors:

OpenAI (or any other frontier lab) is just a customer.

It is not the entire Microsoft bull case.

If OpenAI wins, Microsoft participates.

If Anthropic wins, Microsoft hosts it.

If open-weight models win, Microsoft distributes them through copilot.

If Microsoft’s own MAI models win, it captures more of the margin.

Microsoft owns the toll road.

The models are the cars.

We have been bullish on Microsoft since the start of the year, and have discussed it in this newsletter multiple times, including last week.

That view has finally paid off. Microsoft’s returns are now 3x higher than SPY over the last three months.

The road is becoming more difficult for OpenAI.

The company has now cut GPT-5.6 Luna input pricing by roughly 80%, to $0.20 per million tokens.

It also reduced GPT-5.6 Terra pricing by around 20%, to $2 per million input tokens.

OpenAI is in a difficult spot.

Competition is increasing, and they are losing market share in a market, where customers are increasingly looking to control their spend. 

Enterprises are beginning to use cheaper models for routine workflows and reserve frontier models for the small number of tasks where the additional capability matters.

Microsoft described this as using “the right model for the right job.”

OpenAI can respond by cutting prices and driving more usage.

But usage must grow faster than prices fall.

Otherwise, the company is spending more on compute to generate less revenue per token.

That becomes particularly important as OpenAI moves toward an IPO.

Public-market investors will not value the company only on user growth or model benchmarks.

They will ask harder questions:

How durable is pricing?

What are the margins?

How much capital is required to support growth?

And why should OpenAI receive a similar valuation to Anthropic when Anthropic is reportedly generating roughly 50% more revenue?

That is the tension.

I did a FSD stream this week on OpenAI’s financial troubles, and how Nvidia is the lender of the last resort. Watch it here.

Meta also reported this week, and the reaction was almost routine. It dipped 7% despite reporting 28% revenue growth. 

Revenue reached $60.8 billion, up 28% YoY and ahead of the $60.2 billion expected.

Meta’s core was never the problem, and it was solid this quarter as well.

Advertising revenue increased 27% to $59.4 billion.

Ad impressions rose 14%. The average price per ad increased 12%.

And Meta’s Advantage+ AI advertising products reached a $75 billion annual revenue run rate.

Zuckerberg’s message: “These AI investments are paying off.”

The market had a problem with the bill for these investments.

Capital expenditures reached $31.1 billion in the quarter.

Free cash flow fell to just $784 million.

For 2026, Meta narrowed its CapEx outlook to $130 billion to $145 billion (high end of its range).

When analysts asked about 2027, management declined to provide a number.

Susan Li (CFO) said: “We aren’t providing a specific outlook for 2027 CapEx at this time. Infrastructure planning remains highly dynamic.”

Investors knew what that meant. Capex is not going down anytime sooner — Hence, the stock reaction.

Management is explicitly maximizing capacity across 2026 and 2027, while securing land and power for further expansion beyond that.

Meta says it has plenty of profitable uses for the additional compute. Li argued that its models, consumer products and enterprise offerings will be the “best and highest ROI use of our infrastructure.”

But, the cash goes out today.

The returns arrive later.

Nobody likes the uncertainty.

Meta now trades at roughly 17.7 times forward earnings, near the lower end of its historical valuation range.

At this valuation, the stock is becoming interesting.

But we are not entering yet. I think investors could do well… And if they announce a contract with a frontier lab, the stock jumps 10 to 15% instantly.

But, the opportunity cost is higher in owning cloud businesses directly. It’s a tough call.

The main difference between Google, Microsoft and Amazon is they have a cloud business already. I can show you the ROIC. With Meta, I can show you the capex but the ROIC is speculative.

Apple reported this week and fell 7%.

I noted last week that the stock had returned to rarefied valuation territory, levels we had not seen since the 2021 cycle.

At that valuation, Apple did not need bad news to fall.

It just needed news.

The headline results were actually strong.

Revenue rose 16% to a June-quarter record of $109.4 billion. iPhone revenue increased 22%, Mac grew 29% and Services reached a record $30.7 billion. 

Earnings per share rose 29% to $2.02.

But, the outlook exposed the issue.

Apple guided to stronger foreign-exchange headwinds and worsening supply constraints across the iPhone, Mac and iPad.

Memory costs are also starting to bite.

Tim Cook said Apple paid significantly more for memory in June, expects to pay even more in September and sees market pricing continuing to rise beyond that.

He described the environment as:

“A 100-year flood on the memory pricing.”

Apple has already responded by raising some product prices. 

That leaves two outcomes: absorb the cost through lower margins or ask customers to pay more and risk slowing the upgrade cycle.

The gross-margin guidance made the pressure visible. 

Excluding tariff refunds, management expects margins to fall from roughly 48.1% to 46.5%, with more than the entire decline explained by higher memory costs before partial offsets.

The quarter was not bad.

The valuation simply left no room for anything less than perfect.

July was the month of rotation.

The stocks that had carried the market during the first half were sold aggressively, while many of the names left behind finally caught a bid.

The 20 best-performing S&P 500 stocks during the first half of the year declined 25.3% in July.

All 20 finished the month lower.

Meanwhile, the 20 worst performers from the first half gained an average of 22.7% in July, with 17 of the 20 moving higher.

The most crowded winners were used as a source of funds, while capital rotated into areas where valuations and expectations were lower.

The rotation was violent because the positioning was extreme.

But, that reset has (statistically) created a healthier setup for the broader market.

The Nasdaq 100 recently moved from a three-month low to posting its strongest one-day rally in three months on consecutive sessions while remaining above its 200-day moving average.

That combination has occurred only 14 times since 1985.

And, monthly results are strong after the trigger.

Three months after the signal, the Nasdaq 100 gains an average of 7.5%, with positive returns 86% of the time.

Six months later, the average gain is 13.8%.

At nine and 12 months, the market was higher in 93% of cases, with average returns of 19.7% and 23.1%, respectively.

History does not guarantee the next move.

But the pattern makes sense.

The recent weakness has also made valuations more reasonable.

The S&P 500 now trades at roughly 20.6 times forward earnings, around 15% to 20% below its recent valuation highs.

Semiconductors have seen an even larger reset.

SMH’s forward multiple is roughly 30% below its highs.

The earnings outlook does not need to accelerate from here for stocks to work.

The market simply needs earnings to hold while expectations stop falling.

July cleared out leverage.

It punished crowded positioning.

It narrowed the valuation gap between the winners and the rest of the market.

And it brought the broader indexes back toward levels where future returns are easier to justify.

The correction may not be completely over.

But the conditions for a recovery are improving.

The market has fewer tourists.

Valuations are lower.

And a large part of the excess has already been removed.

Sam Altman argues that as AI models become cheaper and more interchangeable, the lasting value will accrue to companies that control compute infrastructure, embedded workflows, and trusted brands.

View Curation

Dan Ives argues that agentic AI will dramatically expand the corporate attack surface, driving a major cybersecurity spending cycle led by platforms such as CrowdStrike and Palo Alto Networks.

View Curation 

Akshay Nathan explains how AI is removing routine work, collapsing traditional job boundaries, and making adaptability across functions the defining skill of the AI era.

View Curation

For more memes, follow Lumida Memes

Not Subscribed Yet? Don’t miss out on future insights—subscribe to the newsletter now!

For real-time updates, follow us on: 

X | Telegram | Youtube | TikTok | News | Ram’s X | Lumida Health | Lumida Tax | Lumida Memes

As Featured In

Disclaimer: Lumida Wealth Management LLC (‘Lumida”) is located in New York, NY, and is an SEC registered investment adviser. SEC registration does not constitute an endorsement of the firm by the Commission nor does it indicate that the adviser has attained a particular level of skill or ability. Lumida only transact business in those states in which it is registered, or qualifies for an exemption or exclusion from registration requirements. Any direct communication by Lumida with a prospective client will be conducted by a representative that is either registered or qualifies for an exemption or exclusion from registration in the state where the prospective client resides.

The information in this material has been obtained from sources believed to be reliable. While all reasonable care has been taken to ensure that the facts stated in this material are accurate and that the forecasts, opinions and expectations contained herein are fair and reasonable, Lumida, Inc. and Lumida Wealth Management LLC (collectively Lumida) make no representations or warranties whatsoever the completeness or accuracy of the material provided, except with respect to any disclosures relative to Lumida. Accordingly, no reliance should be placed on the accuracy, fairness or completeness of the information contained in this material. Any data discrepancies in this material could be the result of different calculations and/or adjustments. Lumida accepts no liability whatsoever for any loss arising from any use of this material or its contents, and neither Lumida nor any of its respective directors, officers or employees, shall be in any way responsible for the contents hereof, apart from the liabilities and responsibilities that may be imposed on them by the relevant regulatory authority in the jurisdiction in question, or the regulatory regime thereunder. Opinions,forecasts or projections contained in this material represent Lumida’s current opinions or judgment as of the day of the material only and are therefore subject to change without notice. Periodic updates may be provided on companies/industries based on company-specific developments or announcements, market conditions or any other publicly available information. There can be no assurance that future results or events will be consistent with any such opinions, forecasts or projections, which represent only one possible outcome. Furthermore, such opinions, forecasts or projections are subject to certain risks, uncertainties and assumptions that have not been verified, and future actual results or events could differ materially. The value of, or income from, any investments referred to in this material may fluctuate and/or be affected by changes in exchange rates. All pricing is indicative as of the close of market for the securities discussed, unless otherwise stated. Past performance is not indicative of future results. Accordingly, investors may receive back less than originally invested. This material is not intended as an offer or solicitation for the purchase or sale of any financial instrument. The opinions and recommendations herein do not take into account individual client circumstances, objectives, or needs and are not intended as recommendations of particular securities, financial instruments or strategies to particular clients. The recipients of this material must make their own independent decisions regarding any securities or financial instruments mentioned herein and should seek advice from such independent financial, legal, tax or other adviser as they deem necessary. Lumida may trade as a principal on the basis of its views and research, and it may also engage in transactions for its own account or for its clients’ accounts in a manner inconsistent with the views taken in this material, and Lumida is under no obligation to ensure that such other communication is brought to the attention of any recipient of this material. Others within Lumida may take views that are inconsistent with those taken in this material. Employees of Lumida not involved in the preparation of this material may have investments in the financial instruments or securities (or derivatives of such financial instruments or securities) mentioned in this material and may trade them in ways different from those discussed in this material. This material is not an advertisement for or marketing of any issuer, its products or services, or its securities in any jurisdiction.