Chips on the Table: AMD and Arm Earnings Report
A detailed look at how both companies performed this quarter, what’s driving growth, where the risks are, and how markets responded.
Well well well, the release of this article was delayed by one day because of life and Mr Trump sends things flying once again. The US and China have agreed to a 90 day “truce” in their trade war. The tariff rate has dropped from 125% to 30% whilst China has dropped theirs from 125% to 10%. This is very good news, the reason why I don’t say great is because it is only for 90 days and the effect on GDP may have already materialised, since it’s a lagging factor. But I won’t be a pessimist for now, it’s definitely good news. And as Q1 earnings season winds down, I think it’s been pretty solid overall. Demand has remained robust, outlooks have more or less held up and margins remain solid. Of course, it hasn’t been great news for everyone, especially not for one of the names I’ll be covering today. But let’s start on a positive note.
AMD
For an intro to AMD, check out my post from the previous quarter. Since I last wrote about AMD in February, shares have drifted lower, even before the broader selloff following Liberation Day. Part of that is market caution around AI infrastructure exposure. But part of it reflects a gap between expectations and what AMD has been able to prove. This quarter was a step forward, but not without complications.
AMD reported Q1 revenue of $7.44bn, up 36% YoY, with strength coming from Data Centre and Client segments. Data Centre revenue rose 57% YoY to $3.7bn, as the MI300 GPU ramp accelerated alongside continued EPYC CPU demand. Client revenue grew 68% YoY to $2.3bn, a strong bounce back from the cyclical bottom we saw in early 2024. Gross margin came in at 50% on a GAAP basis, or 54% non-GAAP. Adjusted EPS was $0.96, up 55% YoY and ahead of consensus.
But those numbers don’t tell the whole story. AMD took an $800mn inventory charge tied to unsold chips affected by new US export restrictions to China. That charge will be reflected in Q2 results, and if fully recognised, it will drag gross margin to just 43%. AMD’s Q2 revenue guidance of $7.1–7.7bn implies flattish sequential growth at the midpoint. Excluding the charge, margin would remain closer to 54%, but the fact that AMD is having to spell out both versions shows how sensitive results have become to geopolitical developments.
The MI300 ramp is perhaps the strongest signal of progress. Management now expects over $4bn in data centre GPU revenue this year, double what was forecast just last quarter. That reflects stronger demand, particularly from hyperscalers like Oracle, and smoother execution. ROCm 6.4, AMD’s updated AI software stack, is out and they are now shipping inferencing and training containers every two weeks to streamline deployment. The MI300X design, AMD’s flagship AI accelerator, is already securing large-scale deployments from hyperscalers. But AMD still isn’t providing a GPU-specific revenue figure or customer-level breakdowns. Which is understandably frustrating to investors. It makes it harder to judge how much of the $3.7bn in Data Centre revenue is GPU versus CPU, or how repeatable these wins really are. For context, Nvidia breaks this down in detail every quarter.
Moreover, ZT Systems was officially acquired at the end of March. They specialise in designing and assembling full server racks made up of CPUs, GPUs, memory, networking hardware, power and cooling systems. These are complete systems built for hyperscalers and cloud providers who need to deploy thousands of AI servers quickly and efficiently. ZT doesn’t just ship components. It builds out full infrastructure ready to be installed directly into data centres. That capability has become increasingly important as AI workloads scale and customers demand faster deployment with minimal configuration. For AMD, this fills a critical gap. Until now, it has sold the core parts:EPYC CPUs, Instinct GPUs, SmartNICs. But has lacked the ability to combine them into full-stack systems. Nvidia already offers this through its DGX platform and has been winning large deals in part because of it. By bringing in ZT’s design and integration teams, AMD can now build and deliver complete AI compute systems using its own silicon and software stack. That includes ROCm for the software layer and a full AMD-controlled solution from chip to cabinet. It simplifies the buying process for customers and allows AMD to bid on infrastructure deals it previously could not. The US-based manufacturing side of ZT is being sold off, so this was clearly about acquiring design expertise and customer access rather than physical production. In short, AMD has gained the ability to deliver what hyperscalers are increasingly asking for: fully built, ready-to-run AI infrastructure.
Client strength this quarter was real, but also straightforward. The 68% YoY gain reflects restocking and weak prior-year comps. Ryzen 8000 series adoption is going well, and AMD is well-positioned for the AI PC cycle, especially with its Ryzen AI Max+ platform. But this is not yet a growth engine. Gaming was flat and Embedded fell 3% YoY, with commentary pointing to soft industrial demand. These segments are no longer dragging, but they are also not driving the narrative.
Balance sheet leverage increased with the ZT deal. AMD issued $1.5bn in debt and $950mn in commercial paper, pushing total debt to $4.2bn from $1.7bn last quarter. Cash and short-term investments rose to $7.3bn. Free cash flow came in at $727mn, up from $379mn a year ago. $749mn of stock was also repurchased. The cash profile is strong, but some investors may be surprised at how quickly strategic capital has been diverted to covering export disruptions and working capital shifts. I think this reflects a subtle shift in priorities. The ZT acquisition and MI300 ramp were supposed to be forward-leaning investments, but now that same capital is being used to stabilise operations in the face of policy risk. That is not a sign of weakness, but it does show how geopolitics can force even well-capitalised firms to play defence just as they are gearing up to scale. So, where does this leave AMD going forward?
Outlook
The outlook is mixed. AI momentum is real and accelerating, and the MI300 ramp is happening faster than expected. Demand is clearly not the issue. But visibility is. There is still no GPU-specific revenue disclosure, no clarity on which hyperscalers are driving sales, and very little on pricing or margin structure for the AI business. Without that level of transparency, it becomes difficult to benchmark AMD’s position against Nvidia or make a case for meaningful multiple expansion. AMD has come a long way. The MI300 line is resonating with customers, the ZT acquisition adds credibility at the system level, and ROCm is progressing. But investors want more than a roadmap. They want proof. They want details. If AMD wants to be seen as a genuine leader in AI infrastructure, it must show exactly how that opportunity is being converted into revenue, share gains and durable margin. Until then, it will continue to execute in the shadow of a louder, more visible and way better competitor.
Lisa Su reaffirmed AMD’s full-year outlook for strong double-digit revenue growth, despite the $1.5bn hit from new export restrictions. They expect this headwind to be largely absorbed in Q2 and Q3, with minimal impact in Q4. She noted that MI308 shipments to China carried lower margins, so their removal improves the overall product mix going into the second half. AMD expects gross margins to recover as newer GPUs and enterprise CPUs scale, with Lisa emphasising that data centre remains the primary growth driver. She framed the regulatory challenges as manageable and temporary, saying they are more than offset by the momentum in AI infrastructure. MI350 remains on track to ramp mid-year, and AMD is already working with customers to co-design MI400 rack-scale systems. Su described customer engagement as deepening, not just in cloud and enterprise, but also in sovereign AI deployments. AMD is leaning into these changes with increased investment in go-to-market and platform integration, positioning itself to scale through volatility rather than around it.
Price Action
As mentioned, AMD 0.00%↑ has been in a strong downtrend and it still is in one. Despite this, there has been a strong bounce from the $80 mark and if AMD breaks the most recent high, just under $120, that would invalidate the downtrend as in downtrends, you get lower lows and lower highs. If a higher high is made, = downtrend is broken. This doesn’t mean much when macro vol kicks in, but for no,w AMD could creep up further if a new high is made, hitting ~$125.
Onto the next name.
ARM
Arm Holdings were founded in Cambridge back in 1990 and are now majority-owned by SoftBank. They only returned to public markets late 2023, but their architecture has been everywhere for decades. Arm doesn’t make chips, instead, they design the instruction set architecture that defines how software talks to hardware. It’s the foundation layer beneath almost every smartphone chip in the world. Chipmakers license it, build their own designs on top, and pay Arm a fee for every unit sold. This model gives them massive reach without having to manufacture anything. As computing spreads across devices, workloads and geographies, that foundational layer only becomes more valuable.
Arm reports in two segments: licensing revenue and royalty revenue. Licensing includes upfront payments for access to IP and tends to be lumpy depending on the timing and size of deals. Royalty revenue is more stable and tracks chip volumes, but it also improves as newer architectures like Armv9 gain traction. Both are leveraged to long-term industry trends, but timing swings make quarter-to-quarter interpretation more nuanced.
For Arm this was their Q4 for FY25 (this is still referring to the first 3 months of the year). Q4 revenue was $1.24bn, up 34% YoY. This included record licensing revenue of $634mn, up 53% YoY, driven by a mix of high-value custom deals, stronger AI-related demand and a new agreement with the Malaysian government focused on building a national AI ecosystem. Royalty revenue came in at $607mn, up 18% YoY, helped by the broader adoption of Armv9-based chips and growing deployments of compute subsystems across cloud and mobile. Armv9 now makes up over 30% of total royalty shipments, up from 25% last quarter, and carries a higher rate per chip. Smartphone royalties rose more than 30% YoY despite just 2% unit growth industry-wide, showing the benefit of pricing power and architectural mix. Infrastructure royalties also accelerated, lifted by early contributions from hyperscalers deploying Grace Blackwell and Axion. Adjusted EPS came in at 55 cents, beating the 52 cent consensus and landing at the top end of guidance. Non-GAAP gross margin was 98%, flat from last quarter and a reminder of the economics Arm can command at scale. Operating expenses were $566mn, slightly below expectations due to some spend shifting into Q1. Non-GAAP operating income reached $655mn, with a 53% operating margin. Annualised contract value rose 15% YoY to 1.37bn dollars, above management’s typical mid to high single-digit target. That figure smooths out licensing volatility and suggests a healthy trajectory underneath the quarterly noise.
Free cash flow was $163mn, down YoY due to higher capital investment and the reversal of a $573mn working capital benefit last year tied to IPO-related share awards. Arm ended the quarter with $2.1bn in cash and no debt. The balance sheet remains clean, and cash generation is still strong when adjusting for the one-off impact. So far so good right? Well, the markets weren’t pleased regardless. Shares fell nearly 9% after the report… The issue wasn’t execution, it was the outlook. Management declined to give full-year guidance, which is typically worrisome, and instead projected Q2 revenue between $1.0bn and $1.1bn with EPS in the $0.30 to $0.38 range. That implies a sequential drop in both revenue and profitability. Management said this was due to the timing of licensing deals, not underlying demand. That may be true, but when a company is trading on expectations, any lapse in visibility becomes a liability. CFO Jason Child explained that they are receiving less forward-looking visibility from partners than usual, particularly on the royalty side, and that issuing full-year guidance under those conditions would have required an excessively wide range that may not have been helpful.
Price Action
Arm definitely isn’t the prettiest chart to look at. It’s very sideways with wild swings. Of course, because it hasn’t been public for that long, the chart hasn’t had the time to create historical data that can be looked at. Thus, the levels being marked on the chart aren’t as strong as they don’t date back for many years. For me, the next clear level that Arm has to reach is $150, but I don’t think Arm will get there anytime soon. I expect price to consolidate unless a massive headline comes in and sends things swinging. It’s also worth noting that Arm trades at a high premium of 155x (meaning it’s P/E Ratio is ~155). As we’ve seen, in times of turmoil, the first stocks to crash are those trading at a large premium.
Outlook
The long-term story remains intact, but the next few quarters may test patience. Licensing will stay lumpy by nature, and royalty growth, while structurally improving, still depends on end-market volumes that are difficult to predict. Management guided Q2 royalty growth of 25% to 30% YoY, supported by stronger Armv9 adoption, share gains in smartphones, and hyperscaler rollouts in the cloud. They expect infrastructure royalty growth to outpace mobile in the year ahead, driven by ramping contributions from Grace Blackwell, Axion and other custom silicon. Jason Child said royalty visibility is improving but still subject to changes in partner shipment timing, particularly in mobile. On the licensing side, Rene Haas noted strong interest in AI-optimised designs and said pipeline conversion remains healthy, but the company is being measured in how it recognises licensing revenue given variability in deal size and close timing. Without a full-year guide, the market will continue to focus on quarterly consistency, but management made clear they expect royalty momentum to accelerate as new design wins move into production. The opportunity is building, but translating it into a smooth financial cadence is still a work in progress.
Final Thoughts
Q1 earnings have been solid across most of tech. Demand is intact, margins are holding up, and AI infrastructure spend continues to drive the narrative. But what’s becoming clearer is that the market is starting to distinguish between companies that are positioned well and those that can consistently show that position in the numbers. AMD are clearly progressing. The MI300 ramp is ahead of schedule, ZT fills a real capability gap, and client momentum is returning. But without clear disclosure on GPU revenue, customer breakdowns or pricing mechanics, investors are left piecing things together. That limits how much credit the market is willing to give, especially when Nvidia set the standard on visibility. AMD don't need to match that overnight, but they do need to start narrowing the gap.
Arm’s model is structurally attractive. Their IP is everywhere, and royalty leverage only improves as Armv9 scales and more custom silicon moves into production. But being critical to the ecosystem is not the same as being predictable to the market. Royalty lag and deal timing make the model inherently lumpy, which is fine, but the lack of full-year guidance made that volatility harder to overlook. The long-term demand is real. What investors want now is confidence from management and some degree of certainty.
I think from here on out, in the near-term we will enter a risk-on phase with equities doing well. This will be mainly fueled by the tariff relief, with the next big “risk catalyst” coming in June when the Fed may cut rates. This is of course dependent on the data released beforehand.
Thank you for taking the time to read this article, I hope it was insightful! Please share this with friends, family or anyone else who may be interested, as it helps the page grow.
Disclaimer: This article is for informational purposes only and should not be considered financial or investment advice. The views expressed are my own and based on publicly available information, market trends, and personal analysis. Readers should conduct their own research and consult a financial professional before making any investment decisions.