Being Exponential | Luke Lango

Donald Trump’s AI push fuels revolt in Maga heartlands Republicans fear backlash against White House agenda could undermine support in this year’s midterm elections

This, IMO, is the biggest risk facing AI stocks right now. And, indeed, the whole stock market.

Worth monitoring closely...

#NVDA #AMD #META #AMZN #MSFT #GOOGL #AAPL #AVGO #ORCL #TSLA

🔗 www.ft.com/content/0c9...

If that push scales... and we get regulations here... that removes one of the biggest tailwinds fueling the AI infra buildout... and that could cap or even reduce hyperscale AI capex... which would mean that 2026 could indeed be "peak capex" for this cycle.

But there increasingly seems to be pushback against the rapid growth of AI -- from job displacement fears to rising electricity costs because of datacenters -- and that is starting to create a bipartisan push against AI.

A lack of regulation from the Trump administration is currently one of the biggest tailwinds fueling the massive AI infrastructure buildout. Build, baby, build is the mantra.

Excellent article from FT on why AI will become a major hot topic going into the midterms and why politics may actually be the biggest risk to the AI stock bull market.

I like the opportunities presenting themselves in the physical AI supply chain.

2026 hyperscale 5 capex of $710B is a new floor, not the cycle ceiling, especially given the exponential improvements in capability we've seen from ChatGPT, Gemini, and Claude over the last few months.

The spending is working.

So it will continue.

Bubble talk is presently overstated.

#META agreeing to buy a bunch of #NVDA products in a new multi-year partnership which signals we’re still in the early innings of the AI infrastructure buildout.

Humain investing $3B into xAI ahead of SpaceX merger as the AI funding tap remains wide open.

Where do you think all that money is going to go?

Right into data centers. Which is why recent choppiness in AI supply chain stocks looks like a long term opportunity.

Ugly #FVRR results underscore that AI disruption risks are very real. Foundational AI models will increasingly squeeze the middle layer of software. Proprietary data, sensitive workflows, & physical products are moats - but if you have none of those three, you may find yourself in the SaaS graveyard

Graph depicting a downward trend in tech-related information employment from mid-2023 through 2026, showing a significant drop.

If you don't think that AI is killing jobs, then you aren’t paying attention to the data.

And this trend will spread beyond software to finance, sales, hospitality, etc as AI gets better and better.

The chart shows historic divergence between IGV (orange) and NDX (blue) since 2010, highlighting recent significant separation in trends.

This is just… wow. First time ever really that #IGV has diverged so much from #QQQ.

Maybe this time's different

Software is due for a dead cat bounce soon. But long term, the outlook for most software stocks is pretty terrible right now.

Some exceptions to that rule: #PLTR #AXON #IOT #PANW #CRWD

With the 10yr diving down to 4%, I really like the set-up in buying the dip in long duration assets here.

This validates the inference acceleration thesis. Strengthens the case for custom silicon plays like #AVGO, #MRVL, #AMAT, #SNPS, #CDNS, #KLAC, etc. The compute stack is diversifying, and the supply chain enabling that diversification looks attractive.

Exactly what we've been tracking. Hyperscalers are building multi-vendor compute stacks optimized by workload. Training and general inference on GPUs. Latency-sensitive inference on custom silicon.

Codex-Spark delivers 1000+ tokens per second for workflows demanding extremely low latency. OpenAI says GPUs remain foundational for training and cost-effective inference, but this carves out a new tier.

Notable custom silicon progress being made at OpenAI.

OpenAI's new Codex-Spark model is their first built for ultra-fast inference. Runs entirely on Cerebras WSE-3 hardware. Not #NVDA GPUs.

They did it with AlphaFold (solved the 50-year protein-folding problem, then made it free for researchers worldwide).

When you can iterate and cheapen cutting-edge tech faster than competitors, you win.

This is what Google does best. Scale next-gen tech while driving costs down. They're doing it with AI models (cheaper and better than other SoTA models). They're doing it with robotaxis. They're doing it with TPUs (better performance per dollar and per watt every generation).

#GOOGL launched Waymo 6th gen -- 42% fewer sensors, 50%+ lower unit cost to under $20K, better camera performance. Quality up, costs down.

- Dallas<->Houston
- Fort Worth<->El Paso
- El Paso<->Phoenix
- Fort Worth<->Phoenix
- Laredo<->Dallas

Two big self-driving updates today:

Aurora #AUR is expanding autonomous trucking to Phoenix. They're targeting 200 autonomous trucks on the road by EOY and already booked through Q3 2026. Route network now includes:

Chart displays S&P 500 maximum drawdowns with peaks marked by red circles, indicating significant stock declines amid near-high levels.

I'm bullish.

But this is concerning.

When tons of stocks are blowing up but the overall market remains close to highs, that's a pretty bearish signal.

Same thing happened at top of Dot Com Boom.

Lots to like about the CPI report, even if market is flat.

So, inflation stays flat around low to mid 2s, and that's "low enough" to get Warsh to cut 3+ times this year, which should inject some upward momentum into stocks. Plus, look at the 10yr yield. Down below 4.1 now. Lower numbers there will boost risk assets.

Given current trends, I think inflation likely stabilizes in the lower to mid 2s. We’re not going back to 2% flat or lower -- the commodity supercycle is too hot for that. But we're not going to accelerate back to the 3s because the demand just isn't there for that either.

That's a pretty soft January CPI inflation report which suggests to me that the "inflation problem" isn't much of a problem anymore.

Headline inflation rate at 2.4%. Prices only up 0.17% in January. Softest January inflation numbers in 3+ years.

Which means that the indiscriminate selling on Wall Street this week is probably creating some pretty good opportunities in AI supply chain stocks.

#AMAT #ANET #NVDA #AMD #AVGO #VRT #KLAC

Bar graph comparing AI models' performance with Gemini 3 Deep Think leading at 84.6%, followed by others at lower percentages.

These models just keep getting better and better.

Weeks after Claude Opus 4.6 set mind blowing records for AI reasoning, Gemini launches Deep Think which blows Claude Opus 4.6 out of the water.

We are in the AI compounding phase, not an AI bubble.

The AI disruption wave is speeding up and spreading like wildfire. The market is scared to death that AI will obsolete so many jobs, companies, and whole sectors.

And the market is usually ahead of the real world…