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

Markets: Datacenter Cancellation Risks; Black Friday Came Early; Semis Are Cheaper Than Before ChatGPT; The Panic, the Calendar and the Cycle; There Is a Bubble; Remember AppLovin; The Labs’ Biggest Risk

Texas Governor-elect Gina Hinojosa is the most powerful woman in the world.

Whatever she says moves trillions in asset prices.

Democrats trounced Republicans in the midterms. However bad you think the shellacking was, it was worse than even that.

Patrick McHenry, former Chair of the House Financial Services Committee, likened the Blue Wave to the Republican midterm wins of 1994 and 2010.

(Note: I will be asking Patrick McHenry a couple of questions on how he sees things playing out from here on the Monday Bits+Bps podcast at 4:30pm ET.)

Markets expected a Dem sweep per prediction markets Polymarket and Kalshi.

Markets did not expect a thrashing.

What was not priced in by markets was Texas.

Texas went Blue at the Governor, Senator, and Judicial levels.

And Texas has 26% of datacenter capacity. That’s why Gina’s voice matters.

The main question then is: is the DNC sweep really an end to the AI capex, or is it a bump?

The Big Picture

Texas isn’t alone.

The datacenter cancellation risk sits in a handful of states that changed hands — Texas, Ohio, Georgia, Michigan, and Nevada — all states where Dems won.

Virginia is the capital of the data center world, with Texas right behind it, followed by California, Georgia and Illinois.

The US has 57 GW of data center capacity live today.

These states host about 40% of the US datacenter construction pipeline.

And, what is the risk?

The new governors have proposed legislative brakes on datacenters:

1) Gina Hinojosa in Texas campaigned on a moratorium on new data center approvals.

2) Amy Acton in Ohio backs a conditional pause unless developers cover their own utility costs and use union labor.

3) Keisha Lance Bottoms in Georgia has promised an executive order pausing new approvals.

4) Iowa and Nevada’s incoming governors want to end AI tax breaks.

What’s Actually at Risk?

The risk is 2029.

Projects already under construction cover only about 20% of the capacity expected to come online in 2029.

The other 80% still needs a permit, and that’s exactly where the new rules bite.

So, the next question is: what would the new rules mean for the capacity not yet under construction?

I think the best way to understand it is to see what electoral candidates are demanding from datacenters.

Let’s look at Gina Hinojosa — Texas’s Governor-elect.

Texas has the second-largest share of compute capacity, and legislation there will act as a precedent for other states to follow.

Gina wants legislation passed to ‘protect the communities’ that host the datacenters.

(Virginia’s Loudoun County budget now has $1 Bn+ in datacenter tax revenue funding public infrastructure… If that’s what being harmed looks like, most counties would sign up tomorrow.)

This legislation would require that data centers:

(1) pay taxes

» Yes, of course large tech companies should pay taxes. And, many already contribute to their communities where there is effective local governance.

(2) pledge not to raise residents’ electricity rates

» that means more behind-the-grid solutions, which is what datacenters are already doing with nuclear energy and nat gas providers.

(3) protect residents’ access to water

» Texas datacenters touch 0.46% of water. This is fearmongering, not a real issue. And voters don’t rank datacenters as a top issue.

(4) use union labor during construction.

» Mark Zuckerberg is spending over $115M on training blue-collar workers… Big Tech companies will spend on union labor.

What’s the bottom line?

(1) All of the demands on datacenters will be fulfilled.

We expect the legislation for new datacenter builds to be modeled on the original Loudoun County bargain.

(Don’t forget hyperscalers are major contributors to the Democrat Party, and Anthropic has a lot of ex-Biden era staffers. This is 80% political theater, and 20% substance.)

(2) 2026 & 2027 datacenter build is intact, per Texas’s Governor-elect. So, the debate is around 2028. And, there’s plenty of time for policymakers to resolve that.

The Returns Are Too Good to Walk Away From

The hyperscalers are earning returns north of 30% on their invested capital. Their cost of capital is a fraction of that.

That creates a huge cushion.

These projects still make sense if they cost two or three times as much to build.

Double the cost, and a 30% return is still about 15%.

Look at the commitments.

Anthropic’s IPO filing shows nearly $900 billion of compute and infrastructure commitments running out to 2045, some on 20-year take-or-pay contracts.

And, the industry is already adapting.

OpenAI is already working with communities to mitigate the negatives around datacenter construction.

Sarah Friar (CFO, OpenAI) listed what OpenAI offers host communities: a promise never to raise local electricity bills, recycled water, a 300-foot berm in Georgia to block the noise, and free school lunches and a park.

(Imagine getting a politician to honor their commitments!)

Management finds a way.

Nvidia has approached Apollo about converting more than 130 old telecom buildings into AI data centers.

Existing infra means existing power and fewer fights.

Slow It Here, and It Gets Built Somewhere Else

We’ve seen this movie before.

Singapore froze new datacenter approvals from 2019 to 2022. And, Johor, just across the border in Malaysia, became the region’s boom market almost overnight.

Amsterdam paused in 2019, and the demand moved to other European hubs.

Dublin froze grid connections in 2022 and still hasn’t recovered.

In every case, the capex survived.

(So, tech companies have a BATNA, although their existing commitments do cause a one-time hit.)

Projects are already eyeing states that want the investment, like Louisiana, Wyoming, West Virginia and Mississippi, where local support is strongest.

Or the capacity leaves the country, taking the jobs, the tax base and the growth with it.

Anthropic is pursuing projects outside the US, and Meta is weighing overseas sites to get around the new hurdles.

Every site that leaves is a win for another country.

If Democrats squeeze datacenters too hard, they risk making the mistake of the European Union in driving out tech investment to other locales.

We note that Texas’s Governor-elect distanced herself from Progressive labels. She calls herself a populist.

And, our sampling of the newly elected, including Georgia Senator Ossoff, indicates they are more centrist than leftist.

The politicians know this. They understand the threat posed by Chinese AI leadership.

Democrats passed the CHIPS Act.

They understand that semiconductor leadership is a national security priority, and that the buildout expands the tax base.

Which Names Are Exposed?

If you own compute capacity that’s already running, a moratorium makes your asset scarcer.

If you need new permits, it makes your life harder.

The losers are developers concentrated in Texas that still need new builds.

IREN has about three-quarters of its capacity in Texas. That’s the most exposed name in the space.

Debt-funded neoclouds have a second problem. They borrow at much higher rates than the hyperscalers, so every month of delay costs them more.

The clearest winners are the hyperscalers.

Microsoft and Amazon already own the largest fleets of running compute in the world, and both rallied more than 6% in one session this week.

Every delayed project makes the capacity they already have more valuable. And, slower construction means less strain on their cash flows.

The chipmakers should benefit from a second-order effect.

Every permitted data center through 2028 still gets built and still needs Nvidia GPUs made by TSMC.

Every chip Nvidia and TSMC make will continue to get sold. We believe these are compelling bargains once again.

Scarce compute will push spot prices higher, and higher prices mean more demand for chips, not less.

Did the Bubble Burst?

Bubbles burst when earnings estimates start missing and capacity sits idle. There is no idle ‘dark fiber’.

Today, earnings are beating and spot compute prices are at record highs.

Spot prices will go even higher now.

We’ve seen this setup before.

In 2018, Democrats took the House in the middle of a trade war, and the market corrected nearly 20% before a strong rally.

Today, we see semiconductor valuations have reset.

Shield AI agreed to buy Echodyne, a maker of compact radar that detects and tracks drones, in a deal valued at $5.3 billion.

To pay for it, Shield is raising $2.5 billion. That includes $1 billion of new equity at a $28.8 billion valuation.

We entered Shield at a sub-$10 billion valuation in January.

Defense tech is the beneficiary of a multipolar world. 

Ukraine and the Middle East showed that a cheap drone can take out equipment worth millions. 

Every military is now rethinking what it buys. The Pentagon’s budget for drones and counter-drone systems is set to triple.

We’re also seeing some exciting defense tech opportunities in private markets and expect to bring them to our investors soon.

If you are an accredited investor and want to learn the space first, download our defense tech primer here.

It’s a review of the top companies in defense tech.

If you’re an accredited investor or qualified purchaser, join our deals list at Lumida Deals. 

Windows on these deals can be short, so it helps to be on the list before the next one opens.

(Disclosure: The Shield valuation reflects a private funding round and is an unrealized mark, not a realized return. Private investments are speculative, illiquid and can result in total loss of capital. Past performance is not indicative of future results. This is not an offer to buy or sell any security; any offer will be made only to eligible investors through definitive offering documents.)

Since the AI era began, US stocks have rarely been this good a deal.

The S&P 500 now trades at 18.5x forward earnings, the same multiple investors were paying in January 2024.

Compared to last year’s 24x, that’s a 22.6% markdown in just over twelve months.

A falling multiple usually means one of two things: prices crashed, or earnings did.

The former is great for investors because you can get things on sale. And, that’s what happened. 

Over the past six months, the S&P 500 has gone nowhere. It’s up 0.2%. The Nasdaq 100 is down 5.6%, and the Russell 2000 is down 4.6%.

That skepticism is where the opportunity comes from. 

At 18.5x, the market already assumes a good chunk of those earnings never show up. 

If even most of them do, today’s price will look cheap in hindsight.

Compare that with 1999. 

In 1999, multiples rose faster than earnings expectations.

Today, growth expectations are rising and investors are paying less for them.

That’s the opposite of a bubble.

Bull markets end in global euphoria, not local euphoria. People are skeptical of AI. That’s not a bubble. 

A bubble is when your dentist starts giving you stock tips. We don’t see that happening.

The discount is deepest where the earnings growth is strongest.

Semiconductors now trade at 13.8x forward earnings, against 18.3x for the S&P 500. That puts semis at 0.75x the market’s multiple, compared with a five-year average of 1.18x.

Semis are 45% below their January peak multiple of 24.9x.

Look at the left side of the chart. In May 2020, before anyone had heard of ChatGPT, semis bottomed at 0.80x the market multiple.

They’re below that today. 

The market is pricing the AI chip trade cheaper than it priced semis before the AI trade existed.

That’s bananas.

Many hedge funds were liquidated these last few weeks, causing dislocation.

There is some tax-loss related selling pressure too.

The Panic, the Calendar and the Cycle

What to expect ahead?

The setup into year-end has three things going for it: the panic, the calendar and the presidential cycle. 

Each one has a good track record on its own, and this year all three are lining up together.

(1) The Panic

Goldman’s Vol Panic Index surged to 9.57 out of 10 this week, above its 9.5 panic threshold.

The daily jump of about 5 points was the third largest ever. The two-day swing was the largest on record.

That’s the market’s version of a smoke alarm going off because someone burned the toast.

Panic extremes like this have been good entry points.

After the 15 prior signals above 9.5, the S&P 500 rose an average of 4.6% over the next three months, positive 80% of the time. 

Six months out, the average gain was 9.0%. Twelve months out, it was 12.8%, with the index higher 85% of the time.

(2) The Calendar

November seasonality has been the strongest of all months on record. 

Since 2010, November has been the best month of the year for the S&P 500, with an average total return of 2.9%.

This year comes with a twist that has paid before.

September and October both finished red. That has happened three times since 2010: in 2016, 2020 and 2023.

November rose every time. The gains were 3.4%, 10.8% and 8.9%, for an average of 7.7%.

Three years is a small sample, but the logic holds.

Two weak months send investors to the sidelines with cash in hand, and November gives them a reason to come back.

(3) The Presidential Cycle

The second year of a presidential term is usually the worst of the four. We just lived through it.

The year after a midterm has historically been the best.

Since 1938, the S&P 500 has posted gains 95% of the time in the 12 months following a midterm election.

The average return over that stretch is about 14%, nearly triple the roughly 5% of a typical second year.

The reason is simple. 

Markets hate uncertainty more than they hate any particular outcome. Once the votes are counted, investors stop handicapping elections and get back to pricing earnings.

The new setup helps too. 

We now have a Republican president and a Democratic Congress. 

Since 1937, the S&P 500 has averaged a 9.4% annual total return under a Republican president with a divided Congress.

Gridlock isn’t a bug for markets. Neither side can push through anything radical, so policy risk falls. 

What does get done tends to have support from both parties. 

Immigration and trade reform are candidates. So is banning insider trading in Congress.

But it’s not in AI stocks.

It’s in healthcare stocks, which benefited from expectations of a Democratic sweep.

Take a look at the Healthcare ETF (XLV) chart below.

Notably, healthcare was cheap about 6 months ago and had strong relative strength.

But, this safe haven performed well due to midterm risks.

If healthcare is near a local top, that’s bullish for anything that is not defensive.

Back in our August 9 newsletter, we bought AppLovin (APP) after the stock fell about 20% on earnings day.

The market had a long list of worries then: Meta getting into in-app bidding, slower e-commerce growth, and AppLovin’s data compliance issues. 

The stock traded at roughly 17x forward earnings with a PEG below 0.5x. And, it screamed bargain.

But, to be fair, our call was too early. The stock got cheaper before it got right. 

By the eve of its third-quarter results this week, it had slid to about $272, near 13x forward earnings and roughly half its five-year average.

Then the third-quarter numbers came in. 

Revenue grew 55% to $2.18 billion, about 5% ahead of expectations. Net income grew 72% to $1.44 billion.

As a result, the stock jumped 31% on Thursday.

AppLovin had become what Wall Street calls a show-me story. Investors stopped believing the growth until the earnings were in front of them. 

This week, AppLovin showed them.

The pattern is the same each time. The stock falls on a narrative while the earnings keep compounding, so the multiple shrinks. 

Look at Nvidia. It now trades at 15.4x forward earnings, its lowest multiple in at least a decade.

Meanwhile, Nvidia’s growth is accelerating. 

Last quarter, revenue grew 106% to $96.2 billion, its fastest pace in two years. 

Management guided to roughly 70% revenue growth next fiscal year, against consensus of about 45%, and said supply is what’s holding the number back.

The world’s most important AI company is growing faster and trading cheaper than it has in years. 

That’s the same setup AppLovin had at 13x and Atlassian had at its lows.

You don’t often get real earnings growth at a falling multiple. When you do, the market eventually notices.

We have long been talking about how one of the biggest risks for OpenAI and Anthropic is commoditization of LLMs.

Revolut’s AI strategy shows this in action. 

At the start of the year, Revolut’s 2,000 engineers ran on Claude Code and Codex. 

Some were burning $1,000 to $10,000 a month each. The bills kept climbing, and the returns weren’t keeping up.

So Revolut made an open-source model called GLM its default, running on GPUs it rents from Together AI.

By Revolut’s own count, the bill fell to about a tenth of what it paid OpenAI and Anthropic. 

Engineers went from spending caps to unlimited tokens. Coding time fell 60%, and lead time fell 70%.

Revolut still uses Claude and GPT, but only as the exception, for the jobs they do best. 

Everything else runs on open source.

The frontier models went from the house wine to the bottle you open for guests.

We’ve been making this argument for weeks.

Models like DeepSeek cost about 90% less and are good enough for most of the work, and routers make switching easy.

We think Anthropic has the best technical team in the world. They have a major edge over OpenAI since they don’t have a rev share with Microsoft.

The token mix trend is key to watch…

What do we think about Anthropic’s IPO?

To be clear, Anthropic is not walking into its IPO from a position of weakness. 

Its S-1 was strong. I went looking for the other shoe to drop and couldn’t find it, and the market was green in the days after. 

Cash levels are high, and supplier financing is lined up behind them.

But the next test for Anthropic isn’t the IPO. It’s whether customers keep paying frontier prices for work an open-source model can do at a tenth of the cost.

Cheaper models mean more tokens, and more tokens mean more compute.

The labs are fighting over who serves the token. Semis get paid for every single one.

You can now view Lumida Curations in the Lumida Invest App. Curated insights delivered on time. Download the app here.

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 transacts 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 as to 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.