Hock Tan doesn’t do small. The Broadcom CEO, who spent two decades assembling a semiconductor empire through relentless acquisitions, now has his sights set on a number that would have seemed absurd just two years ago: $100 billion in annual revenue. And he’s telling Wall Street he can see a path to get there.
That’s not a typo. One hundred billion dollars.
To put that figure in perspective, Broadcom reported $51.6 billion in revenue for fiscal 2024, itself a record. Tan is essentially promising to nearly double the company’s top line, driven overwhelmingly by the explosive demand for custom artificial intelligence chips and the networking infrastructure that connects them. In a recent report by Yahoo Finance, Tan’s comments during the company’s fiscal second-quarter earnings call in June 2025 crystallized just how aggressively Broadcom is positioning itself at the center of the AI hardware buildout.
The stock market noticed. Broadcom shares surged past a $1 trillion market capitalization after the company disclosed that its AI-related revenue had reached $4.1 billion for the quarter ended May 4, 2025 — a 45% increase from the prior quarter and more than triple the year-ago figure. For the full fiscal year 2025, Broadcom now expects AI revenue alone to hit $13.7 billion, up from a prior estimate of $12 billion. The company simultaneously raised its full-year revenue guidance to approximately $58 billion.
But here’s what really got analysts talking: Tan’s assertion that three major hyperscale customers — widely understood to be Google, Meta, and an unnamed third party, likely Apple or ByteDance — are each building AI clusters comprising one million accelerator chips. Each. That kind of deployment scale is unprecedented and represents hundreds of billions of dollars in infrastructure spending flowing through companies like Broadcom, which designs the custom AI accelerators (known as XPUs) and the networking silicon that stitches these massive clusters together.
“We are very well-positioned,” Tan said during the earnings call, with characteristic understatement for a man whose company has quietly become Nvidia’s most credible competitor in the AI chip space — not by building general-purpose GPUs, but by designing bespoke silicon tailored to each hyperscaler’s specific workloads.
The custom chip business is Broadcom’s secret weapon. While Nvidia dominates the market for off-the-shelf AI training accelerators, the largest cloud companies have grown increasingly interested in designing their own chips — silicon optimized for their particular AI models and data center architectures. Google’s Tensor Processing Units (TPUs), which Broadcom helps design and manufacture, are the most prominent example. But the model is replicating across the industry.
Broadcom currently works with three hyperscale AI chip customers. Tan has indicated that number will grow to eight by fiscal 2027, with each engagement potentially worth billions in annual revenue. The math gets staggering quickly. If each of these customers deploys million-chip clusters, and Broadcom captures a meaningful share of both the accelerator and networking silicon, the $100 billion revenue target starts to look less like fantasy and more like extrapolation.
Wall Street’s reaction has been overwhelmingly bullish. According to coverage from multiple analysts following the earnings report, price targets were raised across the board. Bernstein’s Stacy Rasgon, one of the most respected semiconductor analysts on the Street, called Broadcom’s AI momentum “stunning” and noted that the company’s networking revenue — driven by products like its Tomahawk and Jericho switch chips — is accelerating alongside the custom silicon business. The two revenue streams are deeply intertwined: bigger AI clusters need more sophisticated networking, and Broadcom supplies both.
The company’s VMware acquisition, completed in late 2023 for $69 billion, adds another dimension to the story. Broadcom has been aggressively converting VMware’s customer base from perpetual licenses to subscription models, a transition that initially drew fierce criticism from enterprise customers facing dramatic price increases. But the financial results speak for themselves. Infrastructure software revenue, which includes VMware, hit $6.7 billion in the fiscal second quarter, up 47% year over year. Tan has restructured VMware’s product portfolio, bundling previously à la carte offerings into comprehensive platform subscriptions and pushing hard on private cloud and hybrid cloud solutions that integrate with AI workloads.
Not everyone is comfortable with the pace of change. VMware customers have complained publicly about price hikes of 300% to 1,000%, and some have begun migrating to alternatives like Nutanix or open-source solutions. But Broadcom’s bet is that the largest enterprises — the ones that matter most to its bottom line — will stay because the switching costs are too high and VMware’s technology too deeply embedded in their operations. So far, that bet appears to be paying off.
The broader context matters enormously. AI infrastructure spending is accelerating at a pace that has surprised even the most optimistic forecasters. Microsoft, Google, Amazon, and Meta collectively spent over $200 billion on capital expenditures in 2024, with a significant and growing share directed toward AI data centers. That figure is expected to rise substantially in 2025 and beyond. Every dollar spent on AI compute creates downstream demand for the networking, storage, and software infrastructure that Broadcom provides.
Tan has been explicit about how he sees the revenue trajectory unfolding. The $100 billion target, while not tied to a specific fiscal year, implies a roughly 25% compound annual growth rate from current levels — ambitious but not impossible given the AI spending trajectory. The key assumptions: continued hyperscaler investment in custom AI silicon, expansion from three to eight major chip customers, and sustained growth in networking revenue as cluster sizes scale from hundreds of thousands to millions of chips.
There are risks. Obvious ones.
The AI spending boom could slow if the technology fails to deliver the productivity gains its proponents promise. Hyperscalers could decide to bring more chip design work in-house rather than relying on Broadcom’s design teams. Nvidia could become more competitive in custom silicon through its own partnerships. And the broader macroeconomic environment — trade tensions, tariffs, potential recession — could crimp enterprise IT spending and slow the VMware subscription transition.
Tariffs deserve special mention. The Trump administration’s escalating trade war with China has created uncertainty across the semiconductor supply chain. Broadcom’s chips are manufactured primarily by TSMC in Taiwan, and any disruption to that relationship — whether through tariffs, export controls, or geopolitical conflict — would have serious consequences. Tan acknowledged on the earnings call that the company is monitoring the situation closely but expressed confidence that current demand trends would hold.
Then there’s the competitive question. Marvell Technology, Broadcom’s closest competitor in the custom AI chip market, has been winning its own hyperscale engagements and recently reported strong results. Intel is attempting a comeback in AI accelerators. And startups like Cerebras, Groq, and d-Matrix are pushing novel architectures that could eventually challenge the incumbent players. But Broadcom’s scale, design expertise, and deep customer relationships give it advantages that are difficult to replicate quickly.
The financial profile of the company has transformed remarkably over the past several years. Broadcom’s gross margins now exceed 75%, driven by the high-value nature of its custom silicon and subscription software businesses. Free cash flow generation is enormous — roughly $5 billion per quarter — giving Tan ample resources to invest in R&D, service the debt taken on for the VMware acquisition, and return capital to shareholders through dividends and buybacks.
Broadcom’s stock has roughly tripled since the beginning of 2024. At recent prices above $240 per share, the company trades at approximately 35 times forward earnings — a premium valuation, but one that reflects the market’s belief in Tan’s ability to execute on the AI growth opportunity. For comparison, Nvidia trades at roughly 40 times forward earnings, suggesting that Broadcom still offers relative value if its growth trajectory materializes as management projects.
Institutional investors have taken notice. Broadcom is now a top-ten holding in many large-cap growth and technology funds, and its weighting in the S&P 500 has increased significantly following its market cap crossing the $1 trillion threshold. The stock’s inclusion in discussions about the “Magnificent Seven” — or whatever the market’s current label is for the largest tech companies — reflects its emergence as a first-tier player in the AI infrastructure buildout.
What makes Tan’s leadership distinctive is his discipline. Unlike many tech CEOs who chase shiny objects, Tan has built Broadcom through a repeatable playbook: acquire companies with strong technology and sticky customer relationships, cut costs ruthlessly, focus R&D spending on the highest-return opportunities, and extract maximum value from each product line. The VMware acquisition was the largest and boldest application of this strategy, but the logic was consistent with every deal before it — from Avago’s acquisition of Broadcom itself in 2015 to the purchases of CA Technologies and Symantec’s enterprise security business.
The AI opportunity represents something different, though. It’s organic growth at a scale Tan has never had before. And it’s coming from a part of the business — custom silicon design — that requires deep engineering talent and close collaboration with the world’s most demanding customers. This isn’t about cost-cutting or financial engineering. It’s about technical execution at the highest level.
Broadcom’s networking business deserves more attention than it typically receives. As AI clusters scale to millions of chips, the networking fabric connecting those chips becomes as important as the accelerators themselves. Data must move between chips at enormous speeds with minimal latency, and the networking silicon that enables this is extraordinarily complex to design. Broadcom’s Tomahawk 5 switch chip, capable of 51.2 terabits per second of switching capacity, is the industry’s fastest, and the company is already developing next-generation products for even larger cluster deployments.
The interplay between custom AI chips and networking creates a powerful competitive moat. A hyperscaler designing a million-chip cluster with Broadcom’s XPUs will naturally gravitate toward Broadcom’s networking silicon as well, since the two are designed to work together optimally. This bundling effect — accelerators plus networking plus software — is precisely the kind of platform lock-in that generates sustained, high-margin revenue growth.
So can Broadcom actually get to $100 billion? The arithmetic isn’t unreasonable. If AI revenue grows from $13.7 billion in fiscal 2025 to $40-50 billion by fiscal 2028 or 2029 — plausible if the customer base expands from three to eight and cluster sizes continue to grow — and non-AI semiconductor and software revenue continues its mid-single-digit to low-double-digit growth, the total approaches Tan’s target. It requires a lot to go right. But then again, the past three years have seen a lot go right for Broadcom that few predicted.
Hock Tan has spent his career being underestimated. The Malaysian-born, MIT-educated engineer turned dealmaker was dismissed as a financial engineer when he first began acquiring semiconductor companies in the 2000s. He was dismissed again when he tried — and failed — to acquire Qualcomm in 2018, a deal blocked by the Trump administration on national security grounds. He was dismissed a third time when he paid what many considered an exorbitant price for VMware.
Each time, the skeptics were eventually proven wrong. Whether the $100 billion target proves them wrong again remains to be seen. But anyone betting against Tan at this point should do so with full awareness of his track record — and the trillion-dollar market cap that validates it.
WebProNews is an iEntry Publication