AMD's $4.75B Bond Sale Funds AI Data Center Push
AMD raised $4.75B in its largest-ever bond sale to fund AI data center expansion. Q2 revenue hit $11.54B, up 50% YoY. Here's what traders need to know.
Key Points
- AMD raised $4.75 billion in its largest-ever bond sale — following an earlier $5 billion offering — to fund AI data center infrastructure as Q2 data center revenue more than doubled to $6.7 billion year-over-year.
- CEO Lisa Su's guidance for a further doubling of data center revenue by 2027, paired with Bernstein's active Buy rating and a 27–28x forward earnings multiple well below NVIDIA's, makes AMD the most attractive risk-adjusted play in the AI chip sector right now.
- Traders should watch AMD's MI350 GPU ramp timeline and any guidance update at NVIDIA's upcoming print, which will set the demand context for the entire sector.
AMD raised $4.75 billion in its largest-ever bond sale last week — the second major debt raise in quick succession after a $5 billion offering earlier in August — and the capital deployment strategy is unambiguous: Lisa Su is betting the company's balance sheet on AI data center infrastructure, and Q2 results suggest that bet is already paying off. Total Q2 revenue came in at $11.54 billion, up 50% year-over-year and ahead of the $11.3 billion consensus, with data center revenue more than doubling to $6.7 billion. Adjusted EPS of $1.66 beat the $1.62 estimate. AMD is not chasing NVIDIA — it's building a parallel infrastructure that the hyperscalers are actively buying.
The Revenue Picture and What's Driving It
The 50% year-over-year revenue growth AMD posted for Q2 is not a statistical anomaly driven by an easy comparable — it reflects genuine hardware demand from cloud infrastructure customers deploying AI workloads at scale. Data center revenue doubling to $6.7 billion in a single quarter is the core story. Amazon, Microsoft, and Alphabet aren't writing $600 billion in combined CapEx checks to buy from a single vendor. Procurement diversification is a strategic imperative for every hyperscaler, and AMD's MI300 series has given those customers a credible alternative to NVIDIA's Hopper and Blackwell architectures for specific inference and training workloads.
Lisa Su's guidance for a further doubling of data center revenue by 2027 is the forward number that should anchor how traders think about AMD's trajectory. If Q2 data center came in at $6.7 billion — annualizing at roughly $26.8 billion — a doubling by 2027 implies AMD is targeting north of $13 billion per quarter from data center alone within six to seven quarters. That's not a moonshot; it's a function of the hyperscaler CapEx pipeline, AMD's improving chip-to-chip interconnect story, and the MI350 series entering volume production. Bernstein has AMD rated Buy and Intel at Hold — a bifurcation that reflects the execution gap between the two legacy x86 players.
The bond market's appetite for AMD paper is itself a signal worth parsing. A company raising nearly $10 billion in debt in a single month — at a moment when the 10-year Treasury yield sits at 4.63% and SOFR is 3.62% — is paying a meaningful coupon for that capital. That AMD's management team chose to raise at these rates rather than wait for lower yields implies they see a narrowing window of competitive opportunity. Data center infrastructure is being built now, and the companies that secure wafer allocation and production commitments now will own the installed base that's expensive to rip and replace later.
The MI350 Competitive Angle and Valuation
AMD's MI350 GPU series is the product the competitive thesis hinges on heading into late 2026 and 2027. The MI300 series established AMD's credibility in AI inference; the MI350 is designed to close the performance gap in training workloads where NVIDIA's Blackwell architecture currently holds a meaningful advantage. AMD's move last week to acquire a chip startup that hardwires AI models directly into silicon is a direct response to that gap — if you can bake model-specific optimizations into the chip architecture itself, you reduce the software-layer disadvantage that has historically made NVIDIA's CUDA ecosystem sticky.
The valuation case for AMD is among the most straightforward in the semiconductor sector right now. At 27–28x forward earnings, AMD trades at a material discount to NVIDIA's premium multiple — a gap that reflects NVIDIA's dominant market position but arguably overstates the competitive moat given AMD's accelerating data center penetration. Bank of America named AMD its top CPU pick and NVIDIA its overall sector pick, a distinction that implies BofA sees AMD as the better risk-adjusted entry point for investors who've already missed the bulk of NVIDIA's move. That framing — AMD as the value play within an expensive sector — is gaining traction with institutional allocators.
The University of Oxford BOLD initiative MOU, which sent AMD shares up more than 2% on the day of announcement, is a softer catalyst but worth tracking as a signal of AMD's expanding AI research credibility. Academic partnerships don't move chips on their own, but they do matter for the software ecosystem — researchers who train models on AMD hardware publish results, write libraries, and build toolchains that reduce the CUDA lock-in advantage over time. The Day 0 support for Qwen3.8 27B that pushed AMD up more than 3% reflects the same dynamic: every major AI model that runs natively on AMD ROCm without degradation is a brick in the software moat AMD is trying to build.
What Traders Should Watch Next
The immediate catalyst calendar for AMD runs through NVIDIA's upcoming earnings print. That might sound counterintuitive, but NVIDIA's Q3 guidance will set the demand context for the entire sector — if Jensen Huang guides to another quarter of explosive data center growth, it validates AMD's own 2027 doubling thesis and gives institutional buyers confidence to add AMD exposure at current levels. If NVIDIA's guide disappoints, AMD won't be immune to the sector rotation, regardless of its own execution.
Soros Fund Management's disclosed positions in Micron and Applied Materials, combined with a boosted TSMC stake, frame the broader institutional positioning in semis right now — diversified exposure across the AI supply chain rather than concentration in any single name. Tiger Global's decision to add Intel while trimming Alphabet and Broadcom is a more contrarian move that reflects a specific recovery thesis on INTC, but it doesn't materially alter the AMD narrative. The Bernstein Buy/Hold split between AMD and INTC captures the consensus view accurately: AMD is the aggressor, Intel is the recovery story, and the two are not in the same trade.
For AMD specifically, the level to watch on the MI350 ramp is any production update that surfaces between now and the Q3 earnings call. TSMC's 3nm capacity — the node AMD's most advanced GPUs currently use — is set to increase 20% by year-end, and AMD's ability to secure a proportionate share of that incremental capacity will determine whether the 2027 data center doubling target is credible or aspirational. At $11.54 billion in quarterly revenue and growing, AMD has earned the benefit of the doubt. The bond market just put $4.75 billion behind that judgment.
The Weekly Investor
Daily market analysis for active traders. Free.
