2026 Investment Outlook
Broadly speaking, we stay bullish on the AI-geared infrastructure build-out. With investing and stock selection, it is tempting after a theme works very well to seek out the “next thing”. Not only do we resist that temptation, but we find that, often, it will lead one astray, out into an investment wilderness of less understood and lower quality positions.
As the 4th year of the AI capex buildout begins, we’ll risk cliché and describe ourselves as entering the 4th inning. With the industry having grown capex to a little north of $400b in 2025, the single-biggest tech investment item (and therefore, all of equity investment) is whether one believes in a buildout to $1t within the next few years. If you believe this, it’s preposterous not to be invested in an ecosystem that is going to 2.5x. We’ll also note that each of the past few years (the first, second, and third innings of the AI buildout), industry-wide capex estimates have been wildly too low.
While we are bumping into bottlenecks and power constraints, we expect capex to continue to grow at robust rates and likely stay north of +30%. This will foster lots of investment opportunities. In addition, the inflation outlook has improved, tariff wars aren’t in the daily news cycle, and we’ll likely have a more dovish Fed Chair in 2026; rate cuts will continue apace. Valuations have compressed, and lower rates along with continued industry growth will enable stock returns from both valuation re-rating and earnings compounding.
1) $NVDA ($186) is wildly cheap on a tech relative basis, and on an absolute basis, given industry-leading growth. Moreover, we’ll argue that NVDA has one of the best business models in history (perhaps the best?) with 75% gross margins, operating margins well into the 60s, a defensible moat, software integrated full stack, low capex/high ROIC model, industry-high R&D spend, tremendous free cash flow, and a cash/investment rich balance sheet that is debt-free.
There are two significant NVDA catalysts for 2026. A) the Q1/Q2 release of models and innovations from Blackwell-trained GPU clusters, which we’ll see very soon from xAI (probably first), OpenAI, Meta, and others. B) The release of Vera Rubin in the second half of the year will step-function improve NVDA’s technology lead.
By the spring or early summer, look for confidence to grow in NVDA, achieving $8 in EPS in 2026, and out-year consensus estimates (CY27 / FY28), which have mostly ranged around $7-$8 in EPS, will steadily revise higher towards $10. A sane but conservative mid-20s multiple will price NVDA shares in the $240-$260 range by the summer, with the possibility for a higher multiple as faith in Rubin grows. NVDA trading in-line or at a discount to the NASDAQ is a tremendous entry point for the company with the best business model in tech.
2) $MU ($285) we are memory and storage bulls and willing to say that this is the time that the memory cycle is different. That it soooo has never been different, therein creates the opportunity today. HBM is a critical technology with long lead times and high capital intensity. A smaller competitive set, now whittled down to Micron, SK Hynix, and Samsung, will remain rational with expansion plans. The longer Micron’s earnings don’t show evidence of a “peak” the more investors will value the shares with forward earnings multiples (not P/B). Micron earnings are set to swell to $35, and the stock multiple should be at least 12x. Targeting $420.
3) $SNDK ($237) Ai has come to NAND with surging demand from AI inference applications. With HDD supply constraints and capacity shortages, NAND is at the start of a powerful upcycle. Early last year, SanDisk launched a new product called High-Bandwidth Flash (HBF), which is aimed at giving GPUs fast access to NAND capacity, which will act to accelerate AI training and inference.
SanDisk will ramp to over $20 in EPS, so even though the shares skyrocketed over the past 6 months, we stay long, and look for continued up-revision, a more persistent upcycle, and additional multiple re-rating. Looking for the shares to trade north of $300 this year.
4) $STX ($275) credit to Kerrisdale Capital Management for putting out one of the best and most well-timed pieces of research on Seagate in May 2025. The stock has been a monster. The connection to the AI data center boom and the launch of generational technology transition to heat-assisted magnetic recording (HAMR) was well previewed. Seagate saw an incredible earnings boost, while the shares have rerated. At $275, STX trades at 18x out-year estimates, but with earnings growing by over 30% this year, we stay long. The industry is a rational duopoly with Western Digital ($WDC), which is also very ownable if you want to own all four of the memory/storage names.
5) $VICR ($109) with power as a constraint to datacenter buildout and performance, improved power efficiency is paramount. Datacenters are currently transitioning from traditional 48V/54V power distribution to 800V DC (VDC) to meet the massive power demands of generative AI workloads. This architecture is expected to become the industry benchmark for "AI Factories" starting in 2026–2027. Vicor Corporation benefits from this transition as a uniquely positioned power-architecture firm that develops solutions for the increases in power density and current delivery required by AI compute. VICR has defensible IP, which is generating meaningful royalty income growing in step functions. There have been design wins with GOOGL and AMD and likely with $NVDA in 2026.
Nvidia moved away from Vicor’s solutions in the early innings of the AI buildout based on supply chain risk management (not technology), so Vicor lost the opportunity with Nvidia Hopper and Blackwell. Subsequently, Vicor has ramped its own fab in Andover, Massachusetts, and is better positioned to win Nvidia business with Rubin. Higher fab utilization is leading to higher margins, and while VICR appears rich, there is a cash-rich (debt-free) balance sheet with a share count of just 44m shares. The market cap is $5b, and the company is in a sweet spot to eventually be bought. There is torque in the share price to additional royalty income or a win at Nvidia.
6) META ($660) Zuck is staying the course with capex spending and AI-related research and development. We expect a strong focus on operational efficiency in 2026 to fund AI growth that will be well-received by the market. Blackwell-trained META clusters will produce innovation in the first half of the year, and the shares are down to 22x forward earnings, making META the cheapest name in the Mag7. The mid $600s is a good entry point, and risk-reward in the shares skews positively.
7) AMZN ($230) Amazon has been the dog of the Mag7 in 2025. Although shares were up just 6% in 2025, this is after a 44% year in 2024 and an 81% year in 2023. AMZN shares have de-rated to the lowest valuation in its corporate history at less than 30x forward EPS. AWS growth lags peers as Amazon did not optimize datacenters for AI. This is changing, and AWS's growth accelerated back above 20% last quarter. With a resurgence in growth at AWS and perpetual market share gains at Amazon’s retail businesses, we look for this to be better appreciated in 2026. Amazon has a decade-long supply chain advantage compared to retail peers and trades at approximately a 10-point multiple discount to slower growth WMT and COST, which makes little sense.
8) $NXPI ($217) the analog semiconductor broad-based industry turn (auto, industrial, power, telecom) has been like waiting for Godot. Through the fits and starts, NXP runs its business conservatively and executes well. Moreover, NXP is a great steward of capital. NXPI consistently reduces its share count by a few % a year, boosting earnings growth. NXPI earnings are set to grow 17% based on the 2026 consensus, and at 15.8x forward earnings, the PEG ratio for NXPI is very attractive for a high-quality, market share-gaining analog business model. We like several analog semis for 2026 (ON, SMTC, LSCC, MCHP, MXL), but we’ll highlight NXPI for our outlook.
9) $IREN ($38) there have been some wild price targets out there for IREN, and while hard to model financials, we see competitive advantages relative to other neo clouds and chance for a path to wild upside and shift what was a 2025 speculative long into a 2026 core long. IREN is a profitable Bitcoin mining operation which generates over $600m in annual EBITDA that is transitioning its growth/investment to Ai datacenter buildouts. We like IREN because a) it has secured preferred partner status with Nvidia and has Nvidia Blackwells secured under longer-term contracts (this will matter a lot as Blackwell trained innovation comes out this year) b) IREN generates ongoing cash flow from Bitcoin mining, c) IREN is vertically integrated owning the land, the power generation, and the data center, d) IREN has a power cost advantage of approximately $0.04/kWh vs peers.
Essentially, IREN is a long-term call option on access to Nvidia Blackwell and Rubin when the time comes. There is a current capacity of over 20,000 GPUs with upside to scale to several hundred thousand GPUs, leading to wild price targets. We’ll give IREN a chance to execute this roadmap after doubling since 2025.
10) $CCJ ($91) we’ll round out our list with a long-duration nuclear bellwether, Cameco, out of Saskatoon, Canada. Cameco provides nuclear power across the entire fuel cycle, from mining and milling uranium to refining, conversion, and fabrication. The company has top shelf assets and a 49% Westinghouse to provide reactor services and fuel fabrication. CCJ is the hands-down investible nuclear asset in the western hemisphere. With much of the nuclear industry infrastructure located in Russia, China, and Kazakhstan, we believe that the nuclear industry will be strongly supported in the US. The stock is expensive on an EBITDA basis, but with CCJ the clear leader and largest market cap in nuclear, we’ll own this name through a very long-tailed industry cycle.
Of course what you say is right. $NVDA is around 14x forward earnings. It’s called not being able to admit you are wrong. Burry took a total loss on his NVDA puts. I think were $120 strikes. Question did he short shares this time or continue to buy .05 delta puts indicating no conviction?
@TechnologyPat@Midnight_Captl I define forward as the next calendar year. Which is pretty typical. That’s 2027, and consensus for that is about $12.60. Which means the buyside uses $13 as conservatively assured based on being conservative about beats being slight over the next few quarters.
@Midnight_Captl And until the full stock of cars in developed markets around the world is autonomous. With cars and robots trillions to be spent on silicon, memory, sensors, compute, all of it….
On Bubbles:
Bubbles can be suspected, predicted, or denied, but are only known about, for sure, after the fact.
If there is a bubble today, it's not in stock prices.
Stock prices are nowhere near bubble territory when you look at large caps, the names in the Mag7, and most large caps throughout the market. Sure, there is some froth in areas, but Mag7 sits at normal valuations with some like $META and $NVDA exceptionally cheap.
If there is a bubble today, it's in "the thing itself". Capex spend and Ai. If Ai is real, there is no bubble in stocks.
Contrast this to 1999, where there was a bubble in stocks, but the thing, the internet was transformational and real.
In 2008, the bubble was in the entire economy/housing/credit.
In 2026, it's either in the thing, Ai, or Ai is transformational and real, and there could very well be continued pile into the leaders, which may start to increasingly be bid up on scarcity value (semis). After all, is there anything else in equities to invest in outside of Ai today?
on the $NVDA print:
3rd quarter in a row of accelerating growth. Sales growth of 85% and 18x earnings is unprecedented.
Guide to a 4th quarter of acceleration to 99%.
Accounts Receivable well controlled - grew 130 bps slower than sales growth. Cash flows massive, investment income exploding, $80b buyback and taking the dividend up to 25 cents.
Even at low end of a reasonable: 22x-28x multiple stock will be $264. Expecting this move over the summer.
Trading cautiously because for a few quarters the stock sold off on great results the next day. Best case would be to trade flat and rally into the close tomorrow.
$NVDA preview:
Unique setup for this quarter. Nvidia traded flat for 10-months and broke $25 above the range in the past couple weeks. Normally very tough for follow through, except that Nvidia got exceptionally cheap during the 10-months - profit grew 75% and the stock stayed flat. The stock is still cheap after the $25 or ~14% rally.
If there is a beat/raise in the vicinity of 3-4% this will launch out year earnings towards $12 EPS. The shares should be set up for another $25 extending the rally, just to get to the low-end of valuation ranges.
Staying positioned for this move into the $250-$260 range. Ok, more to come soon...
China is steadily falling behind in Ai advancement compared to pace of progress within the US now. With the US in the catbirds seat, a lack of urgency out of China for progress is surprising. Xi’s issue for longer term. I’m curious how aware he is or if he is being fed China tech propaganda.
China is surpassing Japan in many industries. But one could argue that Japan has stayed relevant globally, in many tech sub-industries, and culturally, and also has the second best slew of distance runners in the world (after east Africa), and Japan would be a much better place to live, on the whole, with better nature, cleaner air, more beautiful mountains, and oceans, and more sophisticated and better tasting food compared to China.
@Midnight_Captl How about Kathy Wood about 90% lower than here....to miss all NVDA is inexcusable as a tech investor.
Dedicated NVDA shorts exposes an investor/entity as a charlatan, pretending to be short for a purpose, but only fraudulently so, or not actually short in a meaningful way.
Not a void. Just the concern is insignificant
Given the capital needs of Ai/technology the winners and large/mega caps are so insurmountably advantaged relative to small caps there is no reason for capital to ever flow to small caps or Russell in an antiquated mean reversion play because innnovation and growth used to be driven by small caps in a past era…
@negligible_cap Nope. Just happened to come out during a surge higher. No highs to even speak about until at least double digit multiples for $MU and all and memory names and $SNDK
Jensen is on the plane. $NVDA is still way too cheap. High conviction to $250.
Amazing that Elon is included after accusing DT of Epstein files. Just wow. Puts $TSLA new highs in play too.