Citi: Broadcom Pullback Is a Buying Opportunity
Miles Bennett
Broadcom dropped over 14% pre-market after earnings, but Citi argues the sell-off rests on a misread — the market fixated on an unchanged old target while missing that AI book-to-bill hit 2.7×, with the bottleneck in capacity, not demand. Current price implies just 20× 2028 earnings.
Did the numbers actually disappoint?
April-quarter revenue came in at $22.2 billion, with adjusted EPS of $2.44 — both slightly above consensus.
AI revenue hit $10.8 billion, up 29% sequentially, also a small beat.
This means → the print itself was fine. The real signal sits in the order book, not the income statement.
So what spooked the market?
Two items triggered the sell-off: July-quarter gross-margin guidance dropped 300 basis points to 74.0%, below the expected 76.8%; and full-year AI revenue guidance stayed at "over $100 billion" — unchanged.
The margin dip is driven by AI custom chips — ASICs, chips designed to a specific customer's specs — rising from 30% to 38% of the mix. That shift is a sign of growth, not weakness.
In plain terms = Broadcom had already rallied hard this year. At elevated valuations, anything short of a blowout was enough to trigger profit-taking. This is a positioning reset, not a fundamental deterioration.
Why isn't Citi worried about margins?
First, operating expenses fell in tandem. The gross-margin hit was nearly fully offset by lower R&D and SG&A costs. Citi's FY2026/2027/2028 EPS estimates moved only ~1%, holding at $11.45 / $18.17 / $25.25.
Second, HBM cost risk is locked in. Broadcom has secured its high-bandwidth memory — HBM, the high-speed memory AI chips require — at fixed prices and will pass any future increases through to customers.
This means → the two biggest margin risks — expenses and raw materials — are already hedged by management, and the market has barely priced this in.
Why is the 2028 picture undervalued?
The market was disappointed by the old "$100 billion" target staying unchanged, but missed the real upgrade: Broadcom's AI demand visibility now extends from "through 2027" to "through 2028."
Citi projects FY2028 AI revenue of $180 billion, total revenue of $229 billion, and EPS of $25.25.
A longer-term lever: content value per GW — a gigawatt being the unit measuring data-center compute scale — runs $10–20 billion, and it keeps rising as compute density, SRAM, CPU, and HBM content per unit increase. In plain terms = even flat GW shipments can still drive revenue growth through higher per-unit value.
How firm is the customer pipeline?
OpenAI has committed to deploying 1.3 GW in 2027, part of a 10 GW plan through 2029. Meta plans 3 GW by end-2028, with the first 1 GW order — covering ASICs and networking — shipping in H2 2027.
AI bookings in the quarter surged to $30 billion against just $11 billion in actual AI revenue — a book-to-bill ratio of roughly 2.7×. The CEO called demand "insatiable."
The bigger structural addition: Broadcom, Apollo, and Blackstone launched the "AI XPV platform," targeting over 20 GW of compute by end-2028, with Apollo committing the first $35 billion. This means → external capital is financing customer compute purchases, bypassing their own capex constraints — opening a demand channel the market has not yet fully priced.
Is the stock expensive at this price?
Citi maintains a Buy rating with a $500 target, based on 20× FY2028 EPS — implying roughly 21% upside from the June 3 close of $413.
Deutsche Bank simultaneously raised its target from $430 to $515, also calling the pullback a buying opportunity.
Citi's scenario analysis: bull case $625 (25× PE, ~51% upside); base case $500; bear case $375 (15× PE, ~9% downside) — a bull-bear spread of 61 percentage points. This reflects how wide the market's disagreement remains on Broadcom's 2028 earnings path.
Content is for reference only, not financial advice.