This Could End the RAM Crisis

Overview

The video argues that the ongoing DRAM/RAM shortage—now approaching a year long—is unlikely to resolve through the traditional “boom/bust” cycle. Historically, RAM prices collapsed after manufacturers overbuilt and oversupplied the market. This time, the video suggests the crisis may instead be shaped by AI-driven demand, contract behavior, and a major shift in global memory competition, especially China’s CXMT—which could influence how and when prices finally ease.

---

Why the shortage persists (and why the usual crash might not come)

• AI and data centers are driving structurally higher memory demand. AI uses large amounts of server DRAM plus HBM (high-bandwidth memory). Since HBM is made from DRAM dies, it competes for many of the same production inputs as conventional RAM.

• Major memory makers prioritized AI/server supply over PCs. They signed long-term contracts, which may prevent near-term demand from collapsing the way it sometimes did in previous cycles.

• Oversupply dynamics may differ this time. The “second exit” to lower prices—manufacturers racing to add capacity and accidentally creating oversupply—might not play out the same way:

  • Factories take years to build, so expansion timing matters.
  • If multiple firms expand together, oversupply can still happen.
  • However, the video claims memory makers have learned from past cycles where aggressive expansion contributed to destructive price crashes, making companies less willing to repeat those outcomes.

---

The “third exit” changes: China’s CXMT as an unexpected new supplier

The video highlights China’s attempt—historically viewed as difficult—to build competitive DRAM capacity despite export controls on advanced chipmaking equipment.

Key constraints

• The biggest blocker is access to advanced lithography tools, especially ASML’s EUV.
• Additional restrictions reportedly apply to immersion DUV equipment as well.

CXMT’s claimed workaround and progress

The video argues CXMT has found a workaround and is increasingly shipping into real consumer brands:

• Reviewers and industry reporting found CXMT DDR5 performance comparable to premium kits at the same speed, with limitations noted (such as some overclocking constraints).
• The video claims vendors including ASUS, HP, Acer, MSI, Gigabyte, and others qualified CXMT memory in notebooks, including through BIOS support.
• It also notes:
  • HP/Asus/Acer shipping CXMT in notebooks (as reported by Nikkei).
  • CXMT’s Shanghai IPO, described as dramatically valuing the company.

---

Market impact: why incumbents still profit even as CXMT rises

The video explains that when AI absorbed much of the available DRAM supply:

• PC makers were pushed to the back of the line, facing extreme prices and shortages.
• When PC makers sought alternatives, CXMT benefited by selling at higher market prices, not merely through subsidized or discounted pricing.

It contrasts incentives between providers:

• Micron / Samsung / SK Hynix must ultimately answer to shareholder expectations and fear repeating oversupply-driven losses.
• CXMT is portrayed as supported by long-loss state funding aimed at long-term national capability, allowing riskier capacity buildout.

---

CXMT’s capacity trajectory and the limits of its technology

The video claims CXMT is expanding aggressively, with several capacity milestones cited:

• Current production described around ~300,000 wafers/month
• Estimates pushing toward ~350,000 by year-end
• Plans that might include additional fabs (e.g., Shanghai and possibly Beijing) potentially exceeding 600,000 wafers/month under certain scenarios
• For comparison, Micron is estimated at about ~375,000 wafers/month by year-end

Limits due to tool access

The video emphasizes that CXMT still lacks the most advanced tooling:

• Without EUV, CXMT relies on older equipment and more passes, which one estimate suggests increases manufacturing cost by roughly +30%.
• EUV-quality output is presented as still superior for incumbents, but the video argues China’s progress matters most because it demonstrates a pathway once thought “impossible” can work at scale.

---

Why China is doing it (three theories)

The video suggests CXMT’s push is likely not only about undercutting prices. It presents three main motives:

1. Memory security If the US restricts/sanctions advanced memory, China needs a domestic “insurance policy.” In the video’s framing, CXMT could become a target for blacklisting.

2. Import substitution China has historically paid heavily for imported chips; building RAM domestically reduces dependence on imports.

3. Strategic leverage via scale (most emphasized) The video argues China may be trying to become a supplier the world depends on—mirroring past industrial strategies (such as solar, batteries, or steel). This could provide leverage without necessarily triggering a direct price war.

---

What it means for prices going forward

• The video concludes that prices will likely come down, but not necessarily back to prior lows from earlier crash conditions.
• Rather than incumbents deliberately flooding the market, the “endgame” may be that China becomes self-sufficient, gradually reducing how much the world’s biggest buyer (China) needs to purchase abroad.
• Net effect: eventual relief, but likely slower and less dramatic than historical “crash” patterns, because incentives to create a collapse are weaker than in earlier cycles.

---

Presenters / contributors

• No clearly identified individual presenter names appear in the provided subtitles.
• The video mentions third-party reviewers/analysts such as Hardware Unboxed and reporters like Reuters and Nikkei, but they are not presented with specific on-screen presenter names in the subtitles.