By Retro News 5, 2026, 7:14 am EST
In this article: [AVGO] [NVDA] [GOOGL]
The artificial intelligence infrastructure boom has entered a new, dizzying phase of capital intensity, and while investors in Big Tech platforms are flinching at the price tag, the “pick and shovel” makers of the semiconductor industry are celebrating a fresh windfall.
Shares of Broadcom (AVGO) and Nvidia (NVDA) were gaining significantly in early trading on Thursday, February 5, 2026, diverging sharply from the stock of their biggest customer, Alphabet (GOOGL). The catalyst was a staggering revelation from Google’s parent company late Wednesday: a plan to spend up to $185 billion on capital expenditures in 2026, a figure that nearly doubles its already historic spending from the previous year.
For the chip makers, this number is not just a statistic; it is a direct revenue pipeline.
Broadcom shares surged 5.6% in premarket trading, erasing losses from earlier in the week. Nvidia rose 3.4%, signaling that fears of a “capital spending plateau” were premature. Meanwhile, Alphabet stock slid nearly 2% as Wall Street digested the short-term margin pressure required to fund this massive build-out.
Here is why Google’s checkbook is opening so wide, and exactly how it fuels the next leg of the rally for Nvidia and Broadcom.
The $185 Billion “Sticker Shock”
During its fourth-quarter earnings call on Wednesday evening, Alphabet delivered what analysts are calling “The Number.” CFO Anat Ashkenazi guided for 2026 capital expenditures (capex) to land between $175 billion and $185 billion. To put that into perspective, the company spent approximately $91.4 billion in 2025—a year that was already considered a period of hyper-investment.
“We are in a relentless innovation cadence,” Alphabet CEO Sundar Pichai told investors, framing the spending not as a choice but as an existential necessity to support the company’s Gemini 3 models and cloud infrastructure.
The breakdown of this spending is the key for semiconductor investors. Ashkenazi specified that roughly 60% of this capital will go directly into “technical infrastructure,” primarily servers equipped with TPUs (Tensor Processing Units) and GPUs (Graphics Processing Units). The remaining 40% is allocated for data center shells, energy infrastructure, and networking equipment.
For the stock market, the reaction was binary. For Alphabet shareholders, the figure represented a terrifying escalation in costs that threatens free cash flow in the near term. But for the supply chain, that $185 billion is pure revenue potential.
“That is an incredible number. We are almost laughing because that number is so good for the Google cohort,” said Ben Reitzes, head of technology research at Melius Research, in a note to clients Thursday morning. “If you are selling the chips, the cables, or the cooling systems, Google just guaranteed your 2026.”
Broadcom: The Custom Silicon King
While Nvidia often grabs the headlines, the biggest relative winner of Google’s announcement is arguably Broadcom. The semiconductor giant has quietly cemented itself as the essential partner in Google’s “custom silicon” strategy.
Unlike Microsoft or Meta, which rely heavily on off-the-shelf Nvidia GPUs, Google has spent over a decade developing its own AI chips, known as Tensor Processing Units (TPUs). These chips are optimized specifically for Google’s internal workloads—Search, YouTube recommendations, and now, the training and inference of the Gemini AI models.
Broadcom provides the intellectual property, design support, and physical manufacturing connections (largely through TSMC) that make these TPUs possible.
“Google is effectively building its own parallel universe of AI compute, separate from the Nvidia standard,” wrote Harlan Sur, a semiconductor analyst at JPMorgan. “Broadcom is the architect of that universe.”
The math for Broadcom is compelling. Analysts estimate that for every TPU Google deploys, Broadcom generates significant revenue not just from the compute processor, but from the specialized networking chips that connect them. Google’s guidance implies a massive ramp in TPU v6 and v7 production to reduce reliance on expensive external hardware.
“Broadcom’s custom ASIC (Application-Specific Integrated Circuit) business is now arguably the most valuable semiconductor franchise outside of Nvidia,” Sur added. “With a $185 billion budget, Google is signaling a massive shift toward internal silicon. That is a direct deposit into Broadcom’s bottom line.”
This dynamic explains why Broadcom outperformed Nvidia in early Thursday trading. The market views the capex hike as a specific endorsement of the custom chip model. As hyperscalers like Google try to control costs, they are increasingly turning to custom designs to get better performance-per-watt—a trend that Broadcom dominates.
Nvidia: The Indispensable Standard
If Google is building its own chips, why is Nvidia rising?
The answer lies in the sheer scale of the demand. Despite its heavy investment in TPUs, Google remains one of Nvidia’s largest customers. The reality of the AI “arms race” is that no single chip architecture can handle the diversity of modern workloads.
While Google uses TPUs for its own internal training and some inference, its Google Cloud Platform (GCP) customers largely demand Nvidia GPUs. Enterprise clients want to train models on the industry-standard CUDA software platform, which runs exclusively on Nvidia hardware. To stay competitive with Amazon Web Services and Microsoft Azure, Google must offer massive clusters of Nvidia’s latest Blackwell and Rubin GPUs.
“It is not a zero-sum game,” said Stacy Rasgon, an analyst at Bernstein. “Google’s spending is so large that it lifts the entire ecosystem. They are buying every H100 and B200 they can get their hands on for the cloud, while simultaneously building TPUs for Search and Gemini. Nvidia wins on the cloud side; Broadcom wins on the internal side.”
Furthermore, the 60% of capex dedicated to servers includes not just the chips themselves but the complex networking gear required to string them together. Nvidia’s InfiniBand and Ethernet networking businesses are likely to see overflow demand from this spending surge.
Market sentiment for Nvidia had softened slightly in recent weeks amid fears that the “law of large numbers” would catch up to AI spending. Google’s $185 billion bombshell effectively nukes that bear case. It confirms that the biggest players in the industry are not tapping the brakes; they are slamming on the accelerator.
The Hyperscaler Arms Race
Google’s announcement does not exist in a vacuum. It is the latest and loudest signal in a synchronized global infrastructure build-out that rivals the industrialization of the electric grid or the build-out of the internet backbone.
Earlier in the earnings season, Meta Platforms (META) stunned investors by raising its own capex guidance to a range of $45-$50 billion. Microsoft (MSFT) and Amazon (AMZN) have similarly indicated that their capital spending will rise in 2026.
Collectively, the “Hyperscalers” are projected to spend over $300 billion on infrastructure in 2026 alone.
“We are witnessing a Prisoner’s Dilemma played out with hundred-billion-dollar balance sheets,” explained Dan Ives of Wedbush Securities. “If Google slows down, they risk losing the AI era to Microsoft and OpenAI. If Microsoft slows down, they lose to Google. The only safe move for them is to spend. The only safe move for investors is to own the companies selling them the shovels.”
This “forced spending” dynamic is what provides a floor for Nvidia and Broadcom stocks. Even if the eventual return on investment (ROI) for AI remains hazy for the software companies, the hardware purchases are happening now.
The Risk Factors: Overbuilding and Margins
Despite the bullish price action for chip stocks, the news carries a warning sign for the broader tech sector. The divergence between Alphabet’s stock decline and its suppliers’ rise highlights a growing tension: Profitability vs. Capability.
Alphabet’s operating margins are under pressure. The depreciation costs associated with $185 billion in assets will weigh on earnings per share (EPS) for years. If AI revenue does not scale as quickly as the infrastructure, Google shareholders could face a “profitless growth” scenario.
“The market is telling us that the transfer of wealth from software platforms to hardware suppliers is still underway,” said Pierre Ferragu of New Street Research. “Google is taking the risk; Broadcom and Nvidia are taking the cash.”
There is also the medium-term risk of overcapacity. If 2026 proves to be the peak of the investment cycle, chip stocks could face a brutal correction in 2027. However, with Google CEO Sundar Pichai explicitly stating that “supply constraints” remain the primary bottleneck, that peak seems to be at least another 12 to 18 months away.
Technical Outlook for the Stocks
Broadcom (AVGO): Thursday’s move pushes AVGO back toward its all-time highs. Technical analysts note that the stock had been consolidating for months. A confirmed breakout above the $320 level could open the door to a run toward $350. The stock trades at a premium, but its dividend growth and monopoly-like position in custom silicon justify the multiple for many institutional holders.
Nvidia (NVDA): Nvidia remains the most crowded trade in the market, but for good reason. The stock has found strong support at its 50-day moving average. Google’s capex news serves as a fundamental validation of its valuation. Analysts will likely be revising their price targets upward, with $160 and $175 emerging as the next key resistance levels.
Alphabet (GOOGL): The stock is likely to remain range-bound. While the core Search business remains a cash cow, the “capex overhang” will act as a ceiling on the stock price until the company can prove that Gemini 3 is generating tens of billions in new revenue, not just consuming tens of billions in servers.
The Bottom Line
Thursday, February 5, 2026, will be remembered as the day the scale of the AI build-out was truly quantified. $185 billion is a number that is difficult to comprehend—it is roughly the GDP of a mid-sized country, deployed by a single corporation in a single year to build synthetic intelligence.
For the broader economy, the implications are complex. But for the semiconductor industry, the signal is crystal clear. The “AI Bubble” is not popping; it is being encased in concrete and silicon.
As Adam Clark writes for Barron’s, the winners of this phase are not necessarily the ones building the smartest bots, but the ones building the brains that power them. Google has just placed the largest hardware order in history, and Broadcom and Nvidia are the ones cashing the check.
Key Takeaways for Investors
- Follow the Capex: Ignore the noise about “AI fatigue.” Focus on the hard dollar amounts being committed by Hyperscalers. As long as capex is rising, chip stocks have a fundamental tailwind.
- Diversify Within Chips: Nvidia is the GPU king, but Broadcom owns the custom silicon niche. Owning both provides a hedge: if Hyperscalers move away from Nvidia to save money, they move toward Broadcom.
- Watch the Margins: For Alphabet, Microsoft, and Meta, the “spend to win” phase is expensive. Expect volatility in the platform stocks while the supplier stocks enjoy the revenue transfer.
Write to Adam Clark at adam.clark@barrons.com
Next Step for You
Would you like me to create a comparative financial table showing the projected 2026 capital expenditures of the “Big 4” Hyperscalers (Google, Microsoft, Meta, Amazon) to help you visualize the scale of this infrastructure build-out?