MRVL, AVGO, HPE Hit 52-Week Highs as AI Trade Broadens

MRVL, AVGO and HPE stocks climbed to 52-week highs on June 2 as investor demand for AI infrastructure intensified. Marvell Technology jumped more than 32% after Nvidia chief Jensen Huang called it the next trillion-dollar company; Broadcom touched a yearly peak of $488.82 ahead of earnings; Hewlett Packard Enterprise ran to an intraday $64.25 after record quarterly results.

The biggest gains landed in the connective layer of the data center: the custom silicon, the optical interconnects and the servers that move data between Nvidia’s graphics chips. The loudest endorsement of that layer came from Nvidia’s own chief executive, on a stage in Taipei, the same week.

Three Highs in One Session

Marvell Technology (MRVL) did the heavy lifting. The stock closed up more than 32% and printed an all-time intraday high of $291.30, a move that pushed its 2026 gain past 225%. Broadcom (AVGO) rose close to 5% to a fresh yearly high of $488.82, a fourth straight green session as traders positioned ahead of its quarterly report. Hewlett Packard Enterprise (HPE) closed up about 19%, its best single day on record, after blowout numbers released the prior afternoon sent the stock to an intraday $64.25.

Three names, one trade. Each sits at a different point in the AI buildout, and each got a separate catalyst within 48 hours.

Marvell and Broadcom design the custom chips that hyperscalers run instead of off-the-shelf GPUs. HPE builds and sells the servers those chips slot into. The wiring between them is where the money moved.

Why Huang Pointed at Marvell

The Marvell move traces to four words. At the Computex conference in Taipei, Jensen Huang shared the stage with Marvell chairman and CEO Matt Murphy and turned to the crowd.

The next trillion-dollar company, ladies and gentlemen.

Huang, chief executive of Nvidia, the dominant supplier of AI training chips, was making a specific argument, not handing out a compliment. As AI computing gets broken into pieces and spread across thousands of accelerators in a single data center, the hardware that ties those pieces together becomes as critical as the accelerators themselves. That hardware, the optical interconnects, the switches, the custom networking silicon, is Marvell’s core business. Coming from the company whose record data-center quarter set the pace for the whole sector, the comment carried weight a sell-side note never could.

Markets priced it fast. Marvell’s valuation swung from roughly $190 billion before the remark to about $235 billion intraday. Stifel lifted its price target to $321 from $230 and kept a Buy rating, with the firm noting Murphy’s keynote was largely in line with what investors already knew but gave Marvell a high-profile platform to broaden acceptance of its data-center positioning, according to TheFly. Retail sentiment on Stocktwits sat at extremely bullish through the session.

The Design Shops Behind the Hyperscalers’ Chips

Strip out the one-day theatrics and a slower shift explains why Marvell and Broadcom keep grinding higher. The cloud giants are designing more of their own AI chips, and they do not do it alone.

Who Owns the Co-Design Market

Together, Broadcom and Marvell control an estimated 95% of the custom AI ASIC co-design market, the engineering work that turns a hyperscaler’s chip specification into manufacturable silicon. An ASIC, or application-specific integrated circuit, is a chip built for one job rather than general computing. Broadcom holds 70% or more of that design-services business; Marvell takes the remaining 20% to 25%. Who they build for tells the story:

• Broadcom: Google’s TPUs, Meta’s MTIA accelerators, plus disclosed programs with OpenAI, Anthropic and Apple.
• Marvell: Amazon’s Trainium and Microsoft’s Maia chips.
• TSMC: the foundry that manufactures the final product for both.

Marvell’s record first-quarter fiscal 2027 results showed why investors keep paying up: revenue hit $2.42 billion, up 28% from a year earlier, with the data-center segment contributing $1.83 billion, or 76% of the total. The company lifted its full-year outlook to about $11.5 billion and its fiscal 2028 target to roughly $16.5 billion.

Why ASIC Growth Is Outrunning GPUs

The macro numbers back the rotation. Research firm TrendForce, a Taiwan-based supply-chain tracker, expects custom AI chips to outgrow merchant GPUs for the first time in 2026.

• 44.6% projected ASIC growth in 2026, against 16.1% for merchant GPUs.
• 27.8% of the AI server market expected to run on ASIC-based systems, the highest share since 2023.
• More than 28% forecast year-over-year growth in total AI server shipments.

GPUs still dominate, at close to 70% of shipments, per TrendForce’s 2026 AI server shipment forecast. The growth, though, is tilting toward the chips the cloud companies design themselves, and toward the two firms that help them do it.

HPE’s Server Backlog Joins the Trade

HPE arrived at its high from a different direction. The company reported fiscal second-quarter results, for the period ended April 30, after the close on June 1, and the stock gapped up the next session. Revenue was $10.7 billion, up 40% from a year earlier, and adjusted earnings of $0.79 a share topped the $0.53 Wall Street expected. Profit more than doubled year over year.

Orders are the part that moved analysts. They more than doubled, leaving HPE with a record backlog, and AI systems orders alone reached $1.8 billion in the quarter. Management raised its full-year revenue growth outlook to a range of 29% to 33%, and lifted its networking growth target to 72% to 75%, citing customers modernizing infrastructure and scaling AI. The figures are laid out in HPE’s fiscal second-quarter press release.

Price targets followed in a wave. Citigroup went to $70 from $39, Morgan Stanley to $71 from $33, and Raymond James to $74 from $29. Our earlier breakdown covered how HPE’s margin mix shifted alongside the AI server surge, the detail that separates a backlog story from a profit story.

A Rally Riding on Few Names

The upside is real. So is the narrowness underneath it. The AI infrastructure trade is broadening within its own lane, from GPUs into interconnect, custom silicon and servers, while the broader market keeps leaning on a short list of stocks. On one recent record close, nine of the S&P 500’s eleven sectors actually fell while AI names carried the index green.

That concentration cuts against holders the moment momentum stalls. Marvell has more than tripled in 2026, which leaves little room for a disappointing print. HPE’s surge rode a single quarter’s guidance raise; the backlog has to convert. And Broadcom ran into its earnings date already up four sessions, the classic setup for a sell-the-news drop even on good numbers.

None of that contradicts the demand. It frames the cost of being late. Buying a 52-week high in a stock up triple digits for the year means underwriting both the buildout and a valuation that already assumes it.

Broadcom’s Numbers Are the Next Test

Broadcom reports fiscal second-quarter results after the close on June 3, and it is the cleanest read on whether the past two sessions priced reality or hope. Guidance pointed to about $22 billion in revenue, up 47% year over year, with AI semiconductor sales near $10.7 billion, implying roughly 140% growth.

The bar is the trajectory. Broadcom’s first-quarter fiscal 2026 report showed AI chip revenue of $8.4 billion, up 106%, on $19.31 billion in total sales. The company spent 2026 stacking deals on top of that base: an expanded Meta arrangement running through 2029, an agreement to manufacture future AI TPUs for Google, and a supporting role in a record $36 billion AI financing deal involving Anthropic and Google.

Broadcom reports after Wednesday’s close. The AI semiconductor line and the guidance attached to it are what the market will trade against everything the rally has already paid for.

Disclaimer: This article is for informational purposes only and is not investment advice. Equities, including semiconductor and AI infrastructure stocks, carry significant risk, and past performance does not guarantee future results. Consult a qualified financial professional before making investment decisions. All figures are accurate as of publication on June 3, 2026.