Alphabet (GOOG) stock entered 2026 with a remarkably tough act to follow.
Its stock soared nearly 66% in 2025, according to Yahoo Finance, with investors buying into its bull case, backed by Google’s growing AI position, robust cloud business, and resilient Search franchise.
It’s still up more than 10% this year, based on Seeking Alpha data, but has lost 9% of its value over the past three months as investor concerns over spending and returns have collided.
Now, though, Morgan Stanley analysts argue that investors might be overlooking a far bigger AI opportunity taking shape inside Alphabet.
Nevertheless, a lot of the concerns this year resurfaced in Alphabet’s Q2 results.
Revenue surged 24% to $119.8 billion, and Google Cloud shot up 82% to $24.8 billion, underscoring tremendous demand for AI infrastructure and services.
Still, Alphabet bumped its 2026 capex forecast to as high as $205 billion, intensifying Wall Street’s already elevated concerns over AI spending and monetization, as I covered in July.
That said, in a note shared with me, Morgan Stanley now sees a far bigger answer emerging.
The firm is doubling down on Alphabet stock, leaning harder into the bull case, with investors remarkably underestimating a growing piece of the company’s AI business, a shift that could become hugely important for its valuation through 2028.
Google’s TPU opportunity is getting much bigger
Morgan Stanley is sticking with its Overweight rating and $400 price target on Alphabet, implying nearly 15% upside from the stock’s Aug. 21 close of $344.82.
At the heart of the bank’s latest note, though, is a sharp reset to how much sales Google could generate by selling its proprietary Tensor Processing Unit, or TPU, systems to customers.
More Tech:
- Anthropic-powered AI model sends shocking message to employee
- Rocket Lab clears 1st hurdle in its biggest satellite deal
- Apple’s $54 billion iPhone machine may be about to break its biggest ritual
Morgan Stanley estimates Alphabet could monetize first-party TPU sales at nearly $27 billion per gigawatt, up from the previous $20 billion estimate, assuming a 30% gross margin versus 20% previously.
The eye-catching revision follows reports that Google could potentially supply about 1 million TPUs, representing 1.3 GW of capacity, for around $35 billion. Alphabet’s expanded custom-silicon relationship with Marvell Technologies (MRVL) also supports the bank’s view that TPU pricing is running hotter than previously assumed.
As a result of that change, you now have some enormous figures.
Morgan Stanley expects Google to sell 0.3 GW of TPU systems in the second half of 2026, 3.2 GW in 2027, and 4.2 GW in 2028. That translates into $84 billion of TPU-related Google Cloud sales in 2027 and $108 billion in 2028, increases of 35% and 37% from previous forecasts.
The TPU ramp-up changes the scale of Cloud. Morgan Stanley bumped its total Google Cloud sales forecasts by 6% for 2027 and 7% for 2028 while raising Cloud EBIT estimates by 2% in both years.
Also read: Alphabet’s stock buybacks: History & investor impact explained
More importantly, the bank now expects Google Cloud to generate 48% of Alphabet-wide EBIT by 2028. That would turn Cloud into perhaps Alphabet’s most dominant profit contributor.
Currently, Google Cloud is generating nearly one-fifth of Alphabet’s total sales. In Q2, Cloud revenue was at $24.8 billion, or 21% of Alphabet’s $119.8 billion total sales, as reported by Seeking Alpha.
However, there is a big trade-off to consider.
TPU hardware carries lower margins than Google’s core cloud services, compelling Morgan Stanley to lower its cloud margin estimates to 37% in 2027 and 40% in 2028.
Still, the bank expects core cloud incremental margins to hover around an impressive 50%, suggesting the underlying business remains tremendously profitable, even as TPU sales scale.
Benjamin Fanjoy/Getty Images
Gemini 4 could be Alphabet’s bigger catalyst
Apart from Google’s massive TPU opportunity, Morgan Stanley sees Gemini 4 as another major catalyst for Alphabet stock.
The bank expects the model to be released in late 2026 or early 2027 and says that returning Google to, or near, the AI frontier could restore investor confidence in its competitive positioning.
It comes at an opportune time.
Google’s flagship Gemini 3.5 Pro was delayed after reportedly falling short of internal performance goals, primarily in coding, while rivals OpenAI and Anthropic continued to push ahead. CEO Sundar Pichai has since hailed Gemini 4 as a substantially larger frontier model, signaling a faster release cadence.
“We’re really focused, putting a lot of effort into Gemini 4,” Pichai said on Alphabet’s Q2 call. “It’s a very ambitious effort.”
In addition, Morgan Stanley highlights Google’s Flash models, prioritizing speed and cost efficiency. That could become increasingly pertinent as enterprises move from AI experimentation to large-scale deployment, where inference costs matter much more.
And there’s plenty of evidence that Google can compete on this front.
Independent testing by Artificial Analysis ranked Gemini 3.7 Flash among the leading models for intelligence, producing nearly 362 output tokens per second, with an input price of $0.75 per million tokens.
At the same time, scale isn’t theoretical anymore. Gemini now has nearly 950 million monthly active users, putting it within striking distance of ChatGPT, according to Reuters.
Google’s big test with Gemini is productization.
The tech giant gains immensely from deploying the service broadly across Search, YouTube, and Google’s other products in ways that produce durable sales growth.
Related: Bank of America raises price targets on 10 software stocks
What should Alphabet investors do now?
For Alphabet investors, the setup has everything to do with execution rather than proving AI demand exists.
Morgan Stanley’s lofty target implies a superb 15% upside, but getting there will require investors to award a healthier premium as Cloud, TPUs, and Gemini become bigger earnings drivers.
The target assumes 24 times average 2027-2028 earnings, compared to Alphabet’s much smaller growth-adjusted premium today.
Morgan Stanley’s own model values Alphabet at 22 times 2028 earnings at $400, compared to around 19 times at recent prices. That means some of the upside requires multiple expansions, not just higher earnings.
From a broader perspective, investors have recently favored the big hyperscalers, including Microsoft (MSFT) and Amazon (AMZN), since their sheer scale, customer relationships, and control over infrastructure offer far more durable advantages than those of highly leveraged AI specialists.
However, the shift toward capital-intensive AI businesses might restrain their valuation multiples, even if earnings remain robust.
Nevertheless, spending remains the pressure point. Alphabet, Amazon, Meta Platforms (META), Microsoft, and Oracle (ORCL) are expected to spend $750 billion on data centers in 2026, according to BigGo Finance, raising the bar for AI monetization and free cash flow growth.
For existing Alphabet stockholders, though, that argues for holding through, despite the spending-driven volatility, instead of treating every capex scare as thesis-breaking. For newer investors, it might be wise to build positions gradually instead of chasing every rally.
The checkpoints are pretty obvious.
Investors need to monitor TPU sales, cloud margins, and backlog, as well as Gemini’s competitive performance and free cash flows. It’s important to note, though, that the downside is real.
Morgan Stanley’s bear case has Google stock dropping to $225 if advertising slows down, AI pressures margins, and spending discipline deteriorates.
Alphabet, therefore, will continue to look attractive if AI spending consistently produces top- and bottom-line growth. If monetization fails to keep up with capex, the same spending fueling the bull case will become the stock’s biggest constraint.
