AMD's July 22 Test, HBM4 Supply Lock & Intel Foundry Earnings

**Jamie Cole** (0:00)
Our Hardware and Chips Daily News.
I'm Jamie Cole. Thanks for joining me. Today, AMD Catalyst Week, can Zen 6 Venice reverse 17% decline?
AMD stock has dropped 17% from its recent peak, and the company has roughly 48 hours to start reversing that with one of the most closely watched product events of the summer. The July 22nd Advancing AI event is the headline this week. AMD is launching Zen 6 Venice, its sixth generation EPIC server processor, built on TSMC's 2nm node. The claimed performance gain is 1.7 times over its predecessor. That's a meaningful jump, and Meta has already confirmed orders for second half 2026 deployment. There's also an MI455X GPU roadmap update expected. The signal here is that AMD isn't just running a chip launch, it's trying to reframe a narrative that's gone cold. The key implication is this, the stock correction isn't purely about AMD fundament, the correction isn't purely about AMD fundamentals. But corrections don't reverse on sentiment alone. The July 22 event needs to deliver guidance specification, particularly on GPU attach rates and average selling prices for the second half of the year. Analyst targets have been raised into the weakness, which creates a flaw, but also sets expectations. Watch the 500 and 150 support level. If Meta's deployment timeline slips even modestly, the buy the rumor risk is real.
While AMD prepares its catalyst, a structural shift in memory supply locked in quietly last month. Nvidia and SK Hynix announced a multi-year HBM4 partnership in June 2026
SK Hynix is set to capture 50-70% of Nvidia's HBM4 orders. Nvidia also requested a six-month acceleration in delivery timelines. That acceleration request is the part worth focusing on.
It tells you Nvidia's Vera Rubin architecture is memory constrained in ways that the existing supply chain wasn't built to handle. HBM4 isn't a premium option here, it's a bottleneck. Samsung and Micron are racing to close the allocation gap. And wafer capacity is projected to roughly double by 2030
The catch is that faster scaling tends to soften pricing, which will eventually congress SK Hynix's margins even as it wins the volume. Supply urgency now, margin pressure later. Intel reports earnings July 23, and the framing from the company will be optimistic. ASML has confirmed that Intel Foundry is now in high-volume manufacturing of Core Ultra 3, branded Panther Lake, using High-NA EUV. That's a genuine milestone. High-NA was prototype territory not long ago. The appointment of Seok-Hee Lee as EVP of Foundry to accelerate packaging monetization adds another headline. Here's the thing. The important distinction is between technical progress and financial progress. High-NA in production is real, but depreciation accumulates fast at these capital intensity levels, and yield learning curves are steep. External orders and utilization rates are what close Foundry losses. Packaging is a smart focus, but investors will want concrete numbers on the loss narrowing timeline, not another 12 months of technology headlines as a substitute for margin improvement.
Nvidia's Rubin generation infrastructure is entering a different kind of competitive conversation. The architecture uses closed-loop liquid cooling that eliminates the water-consuming cooling fans standard in current data center designs. Several municipalities have already moved to restrict or ban large-scale data center construction on water usage grounds. Nvidia is positioning this as an industry standard response to that regulatory pressure. The operational case is clear.
The unanswered questions are cost-premium in adoption timeline. End-to-end liquid-cooled infrastructure isn't cheap to retrofit, and hyperscalers will weigh CAPEXRI carefully. Whether this becomes a genuine competitive moat for Nvidia, or simply raises the baseline expectation for the whole industry is still an open question. Two broader signals round out the week. ASML raised its 2026 sales outlook to between 43 and 45 billion euros, following 9.3 billion euros in second-quarter net sales. Raised guidance at ASML means sustained tool demand, but it also means extended lead times across the industry. Intel, TSMC, and Samsung are all competing for the same equipment.
CAPEX discipline and lead time management are becoming real strategic differentiators. Consider this, Broadcom is trading near record levels on the back of agent sizes with 155 tons of revenue.
The trajectory looks strong, though the relevant watchpoint is where the data center buildout momentum holds through the second half, given regulatory and environmental headwinds starting to affect hyperscaler timelines. The through line this week is straightforward. A MEM needs July 22nd to deliver enough specificity to hold the floor. Intel needs July 23 to show the foundry losses are actually narrowing, not just the technology improving. And the SK Hynix 8BM4 deal is a reminder that memory supply is now a first-order constraint in AI infrastructure scaling, not a secondary consideration. Two things to watch closely. How AMD addresses GDP revenue visibility for the second half, and whether Intel offers any concrete breakeven timeline on the foundry side. Everything else this week is context for those two questions. Thanks for listening. This podcast was built using AI technology. A YesWe production.