$NVDA CEO Jensen Huang on the recent AI stock sell-off:
"We are at the outset of the AI revolutionโฆ We're at the beginning of it, and whatever happened to the stock market, you should be very happy because now you can buy at a discount. Everybody should be very excited."
I couldn't agree more. $SMH +6% today and almost back to all-time highs.
$MSFT unveiled Project Solara, a chip-to-cloud platform for devices built around AI agents.
Think "Mac Mini-style building block," but for agentic devices instead of a general-purpose PC. Developers can create different shells that all share the same underlying stack.
The bet is that if agents become the interface, hardware starts to fragment again. Badges for workers. Desk devices. Wearables. Local AI PCs. More devices that do one thing well, because the agent handles more of the software burden.
Microsoft is coming at this through the enterprise. $NVDA is pushing it through RTX Spark and local AI compute. $META is trying to make wearables the entry point.
Different routes, but all part of the same race. Everyone is trying to define the first everyday AI device that feels bigger than a chatbot.
And for AI infra, more agent-first devices -> more inference, more memory, more sensors, more networking, more cloud orchestration, and more power demand.
The "iPhone moment for agentic AI" is not here yet, but the stack is clearly getting built.
The Information reporting that $NVDA and $GOOGL are considering $INTC as a backup chipmaker.
Compute demand is growing so quickly, and supply chains are becoming so strategically important, that the world's largest AI players can no longer rely on a single manufacturing pathway.
Google has ordered more than 3 million TPUs from Intel for 2028. Nvidia is testing Intel's technology for advanced multi-GPU processors. These are the biggest dogs in the game. If they need help to meet demand, that means demand >>> supply, which means more money flowing into the supply chain.
This also marks a potentially major validation moment for Intel Foundry.
Intel's comeback story has always depended on proving that it can regain process leadership, improve yields, win external customers, and become a credible U.S.-based alternative to TSMC. This report suggests that the process is moving in the right direction.
Execution risks still exist -- yields, cost, timing, and performance remain critical -- but directionally, this is exactly the type of headline that supports the Intel comeback thesis and confirms that AI infrastructure demand remains structurally undersupplied.
Again, just keep on buying those dips.
$NBIS investing ยฃ1.7B to build out capacity in the U.K. with three new deployments of $NVDA infrastructure.
Meanwhile, $AMD announced plans to invest up to ยฃ2B over the next five years in the U.K. to support advanced computing, scientific research, and workforce development across the region.
And $SKM plans to build a gigawatt-scale AI Cloud in South Korea using the U.S. company's DSX platform, with the first AI factory planned to come online in 2027.
The AI buildout continues with no slowdown in sight.
Mizuho raised the price targets on $SNDK $STX $WDC today + highlighted upside in $MU $AVGO in a big, bullish call on memory.
"We continue to see AI as the driving force behind the supply demand imbalance in the memoryย market, as we note increasing demand in 2027/28E could add further pressure to the market. We continue to see 2026/27E DRAM wafer starts up 10%/up 6% y/y, driven mostly by increasing HBM demand, though we note the higher HBM trade ratio limits growth, while we estimate demand growing 27%/24% y/y, respectively. Within NAND, we note eSSD as the key driver of demand, as we estimate 2026/27E total NAND demand growing 18%/18% y/y, while wafer starts are estimated down 5%/up 3% y/y as we see no significant supply online until 2028E."
In other words, memory will remain structurally undersupplied for years.
The music keeps playing, the cycle rolls on, and the stocks will keep working.
Lots of talk about the U.S. govt taking a stake in frontier AI labs like OpenAI, Anthropic, xAI, or others.
My thoughts:
It wouldn't look like outright nationalization. It would more likely resemble a modern industrial-policy partnership. Washington provides something strategically valuable, like procurement contracts, compute access, loan guarantees, energy support, permitting help, national-security partnerships, regulatory clarity, or infrastructure financing. In return, taxpayers receive upside through passive equity, preferred shares, IPO-linked warrants, revenue-sharing rights, or a sovereign wealth fund-style vehicle.
The cleanest version would be warrants. The government does not control the company, does not take board seats, and does not interfere with day-to-day operations, but it earns the right to participate financially if (when) these AI labs become multi-trillion-dollar platforms.
A more aggressive version could involve direct preferred equity or a national AI fund that owns stakes across the frontier model ecosystem.
Either way, the message is unmistakable. Washington increasingly sees frontier AI as strategic national infrastructure, not as a normal software category. More like semiconductors, defense, energy, and aerospace.
That is super bullish for the AI infrastructure trade because frontier AI does not scale on algorithms alone. It scales on GPUs, XPUs, HBM, networking chips, optical interconnects, fiber, servers, storage, cooling systems, power equipment, data centers, nuclear plants, grid upgrades, and massive amounts of electricity.
If the federal government starts de-risking, financing, or politically supporting the AI labs, it is effectively de-risking the capex cycle beneath them.
That directly attacks the bear case that AI spending is too big, too fragile, or too dependent on a handful of hyperscalers. AI infrastructure becomes a public-private national buildout.
The beneficiaries are the picks-and-shovels companies powering the boom: $NVDA, $AMD, $AVGO, $MRVL, $MU, $TSM, $ASML, $VRT, $ETN, $ANET, $GLW, $COHR, $CIEN, $DELL, $CLS, $CRWV, $NBIS, $IREN, $CEG, $VST, $CCJ, $BWXT, $OKLO, etc.
Government equity talk signals that AI is graduating from a private tech boom into a sovereign priority, which makes the long-term AI infrastructure demand curve look much stronger, deeper, and more durable.
Soโฆ yeahโฆ keep on buying those dips, I guess.
The "AI bears" are having a field day today with all the "tokenmaxxing is over" talk, the mixed report from $AVGO, plus the big crash in $CIEN.
The way some of these bears are celebrating, you would think the whole AI infrastructure trade is collapsing right now.
But it's not.
$SMH is down 4% today... After hitting an all-time high yesterday. It's only down 4% from the highs, still up 70% YTD, and still way above its 50-day, 100-day, and 200-day moving averages. All of those moving averages are sloping healthily higher. Still, it's even above the most recent high of $600 from late May.
This is not a trend reversal in any way, shape, or form. It is a necessary, normal, and healthy pullback in a trade that just got too hot.
And it won't turn into a trend reversal, because if you actually look at what Broadcom and Ciena reported, the numbers are consistent with continued robust strength in AI spending.
And the whole "tokenmaxxing" stuff is way overdone. Of course companies blew through their whole budgets with a brand-new tech that was priced on consumption. Now, they reset, add some guardrails, and then reaccelerate spend once the architecture for reasonable spend is in place.
Again, totally necessary, normal, and healthy.
No trend reversal here. Settle down, bears.
$BWXT supplied the TRISO fuel and processed the HALEU feedstock that enabled Antares Nuclear to reach criticality under the DOE program.
This strongly validates BWXT's end-to-end fuel production capability.
The fuel design drew on TRISO specs BWXT developed for Project Pele, the 1.5 MW microreactor it's building for the Army. Decades of work at BWXT's Lynchburg site gave Antares a proven baseline, which shortened the timeline to criticality.
BWXT remains Antares' TRISO supplier going forward, positioning it to support both customer schedules and broader national requirements for advanced reactor fuel.
BWXT is positioned very well as a critical supplier for the next-gen nuclear build-out, which remains in early innings...
$MSTR is back to buying $BTC, which is nice to see, but I really don't think that's the narrative shift we need to get Bitcoin going again.
Clarity Act needs to pass. Even then, we may have to wait for the midterms to see how that shakes out and the halving cycle dynamics to turn bullish in late 2026 / early 2027.
$AMZN + $GLW strike a new multi-year, multi-billion dollar agreement under which Corning will produce optical fiber for data centers and strengthen the U.S. supply chain.
Yet another bullish demand signal for AI infrastructure, and an especially strong demand signal for optics.
The market keeps trying to call "peak AI capex," but hyperscalers AMZN $META $MSFT $GOOGL keep doing the opposite and locking up long-duration supply agreements for the mission-critical components needed to build bigger, faster, denser data centers.
First, the bottleneck was GPUs. Then it became HBM memory, power, cooling, and electrical equipment.
Now, this Amazon-Corning deal suggests another bottleneck is emerging in high-speed connectivity. That means fiber, optical cable, transceivers, and the broader data-center networking stack. That makes sense, because AI clusters are only as powerful as their ability to move enormous amounts of data between chips, servers, racks, campuses, and regions with minimal latency and maximum bandwidth.
As AI models scale and inference workloads explode, optics becomes a central nervous system for the entire AI economy. That is why this announcement is so important for Corning. It's just the latest in a string of hyperscaler commitments suggesting GLW is becoming a strategic arms dealer in the AI buildout.
More broadly, it is bullish for names across the optics and networking complex, including $COHR, $LITE, $CIEN, $ANET, $MRVL, $CRDO, and $AVGO. The bottleneck is shifting toward bandwidth, and Wall Street is just starting to price that in.
Keep buying those dips.
$NVDA demand checks from Wedbush suggest that recent weakness in AI stocks looks less like the start of a fundamental rollover and more like a tactical buying opportunity.
Wedbush says demand for Nvidiaโs Grace Blackwell systems, including the GB300 and B300, remains elevated, with multiple customers indicating those platforms are becoming harder to source.
The bottom-up fundamentals still scream scarcity, urgency, and sustained growth. Demand and AI capex aren't peaking.
So long as that remains true, dips in AI stocks will be buying opportunities.
$SMH $TSM $MU $AMD $INTC $AVGO $KLAC $QCOM $LRCX $ARM $MRVL
The market doesn't like the hot May jobs report, because it means more hawkish Fed behavior going forward.
However, a strong labor market actually strengthens the AI infrastructure thesis by confirming that corporate cash flows, hyperscaler budgets, and enterprise demand remain intact.
Because, at the end of the day, consumers are funding the AI buildout by buying stuff on $AMZN, converting on ads on $META and $GOOGL, and paying $20/mo to OpenAI and Anthropic.
If those consumers lose jobs, their dollars stop flowing, and the big AI spenders have less money to spend.
Today's hot jobs report shows that the jobs are still growing, so the consumers will keep spending. Which means the hyperscalers will keep investing, and the AI boom will persist.
This pullback looks like a buying opportunity.
New Being Exponential episode is live
- SpaceX IPO hits valuation reality wall but long-term bull thesis intact
- Index fast-tracking is rational, not retail grift, given $2T+ scale
- Buy space stock dips at support $RKLB $PL $ASTS $NASA
- $NVDA PC chip launch bullish for $ARM
- Software rally 45% off lows legit technically, but only top 10% survive
- Drone stocks awakening on White House policy shift $RCAT $AVAV $UMAC
- Stagflation weighs broad economy, but AI trade remains unstoppable $SMH
๐ https://t.co/T805dE8ksE
This will turn into a dip buying opportunity, so let's identify the technical levels where support should show up, and this sell-off turns into a bounce.
$SMH at $570-580. Pair of highs from mid-May. Below that, $500 at the 50-day MA + high from April.
$QQQ at $700. Pair of lows from mid-May. Below that, $670. May low + April high + 50-day MA.
Broadly implies another 3-7% move lower. Maybe a 15% gut check to semis in a worst-case scenario.
Nothing to freak out about. Just a normal, healthy, and necessary pullback in a still-alive AI bull market.
The White House is reportedly considering taking equity stakes directly in frontier AI companies, including talks with Sam Altman and OpenAI about a possible stake.
That's pretty bullish for the AI trade.
The White House looks like it is extending its "President's Portfolio" strategy directly into the heart of the AI boom. Over the past year, this admin has increasingly blurred the line between policymaker and strategic investor, taking or looking at stakes in industries it views as critical to national security and long-term economic leadership, including semiconductors, rare earths, and quantum computing. That approach has already been enormously powerful in the market, with government-backed winners like $MP $INTC $USAR $GFS $RGTI $QBTS and many others. Plus, accelerated growth and capital flows in all those industries.
So, in short, this strategy works to accelerate growth. And now, the White House is taking it directly to the centerpiece of the AI boom. Which should, yes, accelerate growth at the center of the boom -- and that acceleration should expand outward to the whole industry.
The more money/support OpenAI has, the more money they'll spend on $NVDA GPUs, and the more the hyperscalers will order $MU memory, $ANET networking, $VRT cooling, and $GEV turbines to stand up all this new compute.
Yes, the talks are preliminary. Yes, the legal framework is unclear. And yes, government ownership could create messy conflicts of interest. But from a market perspective, the signal matters more than the structure right now. It says Washington increasingly views frontier AI less as a speculative bubble to constrain and more as a strategic industrial base to fund, protect, legitimize, and potentially monetize on behalf of the public. That is a huge psychological and fundamental tailwind for the entire AI ecosystem.
If OpenAI, Anthropic, and other frontier labs become de facto national champions, then the infrastructure layer powering them -- Nvidia, AMD, Broadcom, Marvell, TSMC, Arm, Dell, Super Micro, Vertiv, CoreWeave, memory, networking, power, nuclear, and data-center stocks -- becomes even more strategically important. In other words, this is not "peak AI." It is another sign that the AI boom is becoming state-backed capitalism's next great trade.