Cash markets do not share a clock and RWA perps don’t wait for them to agree. August was another record month against the previous report’s published July baseline. The tracked RWA perp market generated $751.9 billion in RWA perp volume, versus $708.3 billion in July, a 6.2% increase. The market also handled a wider range of situations. Memory equities reversed after a long run higher. Seoul paused program trading during a sharp selloff. Moderna doubled after a clinical readout. Different markets, different schedules, one place to keep trading. The month was larger AND broader: more assets, more types of event, and more reasons to trade around the clock. Yet, still a lot of market left to build. Memory Equities Traded in Both Directions July was a one-way memory trade. August brought the other side of it. The tracked memory complex — SNDK, SKHYNIX, SKHY, MU, SNXX, DRAM, and SAMSUNG — generated $327.4 billion, or 43.5% of August volume. Four of the ten most-traded assets came from the group: SNDK ranked first, SKHYNIX third, MU sixth, and SKHY tenth.
On August 4, the memory trade moved higher. SanDisk rose 8%, Micron 6%, and SK hynix 4% after the companies advanced the first Open Compute Project HBF specification. On August 18, it moved lower: Micron fell 5%, SanDisk 6%, Western Digital 7%, and SK hynix 6% as Treasury yields moved higher and investors repriced the trade. The same group drove both the rally and the selloff and kept trading through both. Korea Sold Off; Perps Kept Trading Korean exposure was not a side story. SKHYNIX ranked third, KORU ninth, and SKHY tenth. Together, they generated $117.1 billion, or 15.6% of August volume. On August 19, the KOSPI fell nearly 6% and a Korea selloff triggered a five-minute sell-side sidecar for program trading. After the close, SK hynix announced a 40 trillion won buyback plan, and the KOSPI recovered almost all of the previous session’s loss the next day. MRNA Perps Listed in Hours. Depth Did Not. On August 19, Moderna and Merck reported positive Phase 3 results for their personalised mRNA cancer vaccine in melanoma. Moderna’s stock rose 176.97% in the session. Very few venues had an MRNA market live before the result; TrueCurrent was one. A few more markets followed shortly after the news broke, including TradeXYZ. From its August 19 listing through month-end, MRNA generated $571.6 million and ranked 69th by volume. Its first two sessions produced $213.5 million combined, representing close to 40% of its total monthly volume. The point is access to the event, not the total volume. Traders already have venues for recurring events around the MAG7 and large technology and AI companies, especially earnings. MRNA showed how that can extend to a smaller public company when a one-off clinical result drives attention. With the right risk, and market-data systems, an exchange can make the event tradable quickly, while the news is still relevant and driving volatility. The 24/7 Reference-Price Problem August brought a market-structure question into focus: who produces a usable price when traditional market infrastructure is closed, paused, or has not opened yet? Douro Labs and the Hyperliquid Policy Center brought that question into the SEC’s market-structure process, arguing that the SEC should recognize qualifying independent reference prices for onchain markets where the SIP-derived NBBO is unavailable or does not reflect onchain conditions. The standard they describe rests on direct contributors, a published methodology, transparent publishers, and checks against traditional market data. A separate SEC comment from the Hyperliquid Policy Center and trade[XYZ] used IPOPs — cash-settled pre-IPO perpetuals with no shares, voting rights, or claim on the issuer — as an example of price discovery before a public listing. The CFTC comment process raises a related question for 24/7 futures and perpetuals in energy markets, where the underlying can keep moving after U.S. futures close. All point to the same shift: perps are bringing questions of data provenance, instrument classification and market access into policy discussions. That is directly relevant to RWA markets. These issues are directly relevant to Pyth, whose data infrastructure is used across much of the tracked volume. August in Numbers August closed at $751.9 billion in tracked volume, a 6.2% increase from the previous month.
Asset Class Ranking The market is very much still equity-led with $487.3 billion or 64.8% of the total volume. Commodities followed at $152.3 billion (20.3%), then indices at $103.9 billion (13.8%) and FX at $8.3 billion (1.1%).
Venue Ranking August showcased a reshuffle behind Binance which is head and shoulders above the rest and still growing ($385.6 billion to $437.4 billion). OKX took the 2nd spot as it held its volume above $100 billion and moved from third to second, while Hyperliquid dropped sharply from July’s second-place position to $84.6 billion in August. The RWA perp volume remained top-five concentrated with over 95% of it being traded on Binance, OKX, Hyperliquid, Bitget, and Bybit.
Market-Data Provider Ranking On the data provider and infrastructure front, Pyth remained the undisputed leader with over $715 billion in RWA perp volume secured, representing 96.27% of the total tracked RWA perp volume. One extra percentage point compared to July further solidifying Pyth Pro and Indices as the products powering 24/7 tradfi markets.
Methodology and Sources All volume, listing and provider figures are drawn from Refraction Research and the RWA Markets dashboard, built by @zinnresearch. Volume is notional traded volume across tracked perpetual venues. Volume priced per market data provider attributes each venue-symbol pair to its stated pricing source, weighted by volume. Pairs without a confirmed source are recorded as unverified.
75·ALong
n
news9/8news
August Yielded a Record Month in RWA Perp Volume
Cash markets do not share a clock and RWA perps don’t wait for them to agree.
August was another record month against the previous report’s published July baseline. The tracked RWA perp market generated $751.9 billion in RWA perp volume, versus $708.3 billion in July, a 6.2% increase.
The market also handled a wider range of situations. Memory equities reversed after a long run higher. Seoul paused program trading during a sharp selloff. Moderna doubled after a clinical readout. Different markets, different schedules, one place to keep trading.
The month was larger AND broader: more assets, more types of event, and more reasons to trade around the clock. Yet, still a lot of market left to build.
Memory Equities Traded in Both Directions
July was a one-way memory trade. August brought the other side of it.
The tracked memory complex — SNDK, SKHYNIX, SKHY, MU, SNXX, DRAM, and SAMSUNG — generated $327.4 billion, or 43.5% of August volume. Four of the ten most-traded assets came from the group: SNDK ranked first, SKHYNIX third, MU sixth, and SKHY tenth.
On August 4, the memory trade moved higher. SanDisk rose 8%, Micron 6%, and SK hynix 4% after the companies advanced the first Open Compute Project HBF specification. On August 18, it moved lower: Micron fell 5%, SanDisk 6%, Western Digital 7%, and SK hynix 6% as Treasury yields moved higher and investors repriced the trade.
The same group drove both the rally and the selloff and kept trading through both.
Korea Sold Off; Perps Kept Trading
Korean exposure was not a side story. SKHYNIX ranked third, KORU ninth, and SKHY tenth. Together, they generated $117.1 billion, or 15.6% of August volume.
On August 19, the KOSPI fell nearly 6% and a Korea selloff triggered a five-minute sell-side sidecar for program trading. After the close, SK hynix announced a 40 trillion won buyback plan, and the KOSPI recovered almost all of the previous session’s loss the next day.
MRNA Perps Listed in Hours. Depth Did Not.
On August 19, Moderna and Merck reported positive Phase 3 results for their personalised mRNA cancer vaccine in melanoma. Moderna’s stock rose 176.97% in the session.
Very few venues had an MRNA market live before the result; TrueCurrent was one. A few more markets followed shortly after the news broke, including TradeXYZ.
From its August 19 listing through month-end, MRNA generated $571.6 million and ranked 69th by volume. Its first two sessions produced $213.5 million combined, representing close to 40% of its total monthly volume.
The point is access to the event, not the total volume. Traders already have venues for recurring events around the MAG7 and large technology and AI companies, especially earnings. MRNA showed how that can extend to a smaller public company when a one-off clinical result drives attention. With the right risk, and market-data systems, an exchange can make the event tradable quickly, while the news is still relevant and driving volatility.
The 24/7 Reference-Price Problem
August brought a market-structure question into focus: who produces a usable price when traditional market infrastructure is closed, paused, or has not opened yet?
Douro Labs and the Hyperliquid Policy Center brought that question into the SEC’s market-structure process, arguing that the SEC should recognize qualifying independent reference prices for onchain markets where the SIP-derived NBBO is unavailable or does not reflect onchain conditions. The standard they describe rests on direct contributors, a published methodology, transparent publishers, and checks against traditional market data.
A separate SEC comment from the Hyperliquid Policy Center and trade[XYZ] used IPOPs — cash-settled pre-IPO perpetuals with no shares, voting rights, or claim on the issuer — as an example of price discovery before a public listing. The CFTC comment process raises a related question for 24/7 futures and perpetuals in energy markets, where the underlying can keep moving after U.S. futures close.
All point to the same shift: perps are bringing questions of data provenance, instrument classification and market access into policy discussions. That is directly relevant to RWA markets. These issues are directly relevant to Pyth, whose data infrastructure is used across much of the tracked volume.
August in Numbers
August closed at $751.9 billion in tracked volume, a 6.2% increase from the previous month.
Asset Class Ranking
The market is very much still equity-led with $487.3 billion or 64.8% of the total volume. Commodities followed at $152.3 billion (20.3%), then indices at $103.9 billion (13.8%) and FX at $8.3 billion (1.1%).
Venue Ranking
August showcased a reshuffle behind Binance which is head and shoulders above the rest and still growing ($385.6 billion to $437.4 billion). OKX took the 2nd spot as it held its volume above $100 billion and moved from third to second, while Hyperliquid dropped sharply from July’s second-place position to $84.6 billion in August. The RWA perp volume remained top-five concentrated with over 95% of it being traded on Binance, OKX, Hyperliquid, Bitget, and Bybit.
Market-Data Provider Ranking
On the data provider and infrastructure front, Pyth remained the undisputed leader with over $715 billion in RWA perp volume secured, representing 96.27% of the total tracked RWA perp volume. One extra percentage point compared to July further solidifying Pyth Pro and Indices as the products powering 24/7 tradfi markets.
Methodology and Sources
All volume, listing and provider figures are drawn from Refraction Research and the RWA Markets dashboard, built by @zinnresearch.
Volume is notional traded volume across tracked perpetual venues.
Volume priced per market data provider attributes each venue-symbol pair to its stated pricing source, weighted by volume. Pairs without a confirmed source are recorded as unverified.
$MU $SKHY $DRAM $NVDA $TSM
ABSOLUTELY WILD
South Korea just announced “AI for All”: every citizen gets free, unlimited AI tokens and access to homegrown AI chatbots and AI agents. Unlimited Inference.
Government pays the bill.
Meritz Securities (Korean sell side): GPT-6 Astra launch and implications for the memory stock rebound
On September 3, OpenAI unveiled GPT-6 Astra. As interpretations of this model became a hot topic, semiconductor stocks rebounded on September 4 even as the broader US market fell on rising rates following a jobs surprise, with the DRAM ETF up 6.6% versus the prior day.
Astra's implication is unlikely to be simply whether AGI has been achieved. What drew the most attention was the score of 99.9% on ARC-AGI-3, a benchmark used to judge AGI, a huge improvement over the previous model (Sol at 7.8%). This benchmark is not a knowledge test; it evaluates a model's ability to learn on its own in an abstract environment it has never seen before, and this is where the improvement over the previous model was large. General intelligence, as measured by AAII or Humanity's Last Exam, did not improve much.
The differentiated strength improved in Astra is "the ability to acquire skills that humans learn in an unfamiliar environment as efficiently as a human does." Where AI until now found the answer by pressing this and that 100 times, Astra has started to behave more like a human: observing the phenomenon, inferring the rules, and executing right away.
The key point is "an expanded scope for replacing human intelligence and human work." If existing AI was an AI that told you what to do, Astra is closer to an AI that, given only a goal, uses the computer directly and produces the result all the way to the end. In other words, an easy to use OpenAI model has begun to handle on its own part of the agent orchestration layer that had been the domain of less accessible tools such as OpenClaw. For users, the barrier to entry for AI agents has been lowered, meaning more work can be handed over.
Expansion of AI workloads
Astra naturally also comes with efficiency gains that lower the token cost per task versus the previous model. This is a trend across the AI industry as a whole, and if AI workloads were fixed, demand for AI data centers would have to plunge.
The reason Jevons paradox continues to operate even after token price declines became a trend following the rise of Chinese models is that AI technological progress also expands the workload. What Astra's technological progress means is that where humans used to hand five minute, ten minute, and twenty minute tasks to AI, as AI performance improves and token prices get cheaper there is more to hand over, such as one hour and 24 hour tasks.
Just as news flow about rising GPU rental prices has spread since Astra's arrival, it must be understood that falling AI token prices do not necessarily shrink or slow the AI hardware TAM. Rather, one should recognize that the emergence of a model like Astra can create another inflection point for the AI industry and structurally grow AI demand.
Our understanding is that since early July, as the pace of GPU rental price increases slowed and token prices fell, a long IGV (software) / short SOX (semiconductors) pair trade has persisted in the US. This is because falling token prices were interpreted as positive for software, where tokens are a cost, but negative for infrastructure.
If progress in models like Astra structurally spreads AI workloads and GPU rental prices begin to respond again, the perception that falling token prices are bad news for AI infrastructure companies could weaken (on 9/4 the DRAM ETF rebounded while IGV fell).
As we have argued consistently, the issues accumulating in the AI industry since June (the proliferation of open models, this GPT-6 Astra release, and so on) are, in our interpretation, positive catalysts that generate new demand for AI infrastructure and hardware that did not exist before. We think that in a phase where rates are rising overall and liquidity is becoming scarce, these accumulated positives are not being reflected.
The stock market in September is still uncomfortable with high rates, and within the Korean market there remain hurdles to get through, including digesting a round of earnings estimate cuts driven by the sharp won appreciation before the 3Q26 preview season. There is still discomfort standing in the way of the accumulated positives being reflected in a sustained trend. Overall, we continue to view the market conservatively.
Even so, as emphasized in our September strategy, we believe one should not substantially empty out positions in core AI infrastructure stocks centered on memory. Positive catalysts not reflected in share prices are accumulating. While our baseline is conservative through mid October, one should keep the upside risk open that the trend, led by AI and semiconductor leaders, could turn at any time, even before October.
0·-Neutral
m
meme9/6meme
SK hynix sharply raises share of leading edge DRAM… 1c to become main process early next year
SK hynix is rapidly expanding the production share of its sixth generation 10nm class (1c) DRAM. With 1c DRAM set to serve as the core die stacked into next generation high bandwidth memory (HBM4E), the company appears to be accelerating its process transition. As demand for high value added memory for servers and artificial intelligence (AI) grows, the race with Samsung Electronics and Micron to migrate to finer process nodes is also heating up.
"1c process share expected to exceed 34% within the year"
According to industry sources on the 7th, SK hynix's 1c DRAM share rose from around 10% in the first quarter of this year to the 13% range in the second quarter. It is projected to reach the 24% range in the third quarter and the 34% range in the fourth quarter. In the first quarter of next year, the 1c share is expected to climb to the 35% range, overtaking 1b (33% range) for the first time and becoming the company's main process.
The share of 1b (fifth generation 10nm class) DRAM was found to have peaked at the 43% range in the second quarter of this year and turned downward. As the production shift to 1c gets into full swing, the share of older generation processes is shrinking in sequence. Industry estimates show that, as of the end of the second quarter, Samsung Electronics' 1c share stood in the 16% range and Micron's in the 19% range, somewhat ahead of SK hynix (13% range). However, with SK hynix stepping up the pace of its transition in the second half of this year, it is expected to overtake Samsung Electronics (31% range) on a fourth quarter basis (34% range).
On its second quarter earnings conference call last month, SK hynix said that supply of DRAM built on the sixth generation 10nm class (1c) process had begun in earnest in the second quarter. The company projected that bit growth (the rate of increase in production volume) in the second half of this year would exceed the first half, driven by expanding HBM4 (sixth generation HBM) volumes and rising shipments of 1c based commodity DRAM.
Process transition in preparation for HBM4E performance gains
The battle for HBM4 leadership between Samsung Electronics and SK hynix is also intertwined with the pace of the 1c transition. Samsung Electronics is applying 1c DRAM from the HBM4 stage onward and is touting top tier performance with operating speeds of around 11.7Gbps. SK hynix, by contrast, chose a strategy that prioritizes mass production stability, relying on its proven 1b DRAM and advanced MR-MUF packaging technology, and is applying 1c DRAM as the core HBM die for the first time starting with next generation HBM4E (seventh generation HBM).
Industry observers say this difference in strategy is being reflected in the two companies' market shares. Major research firms including Counterpoint Research project this year's HBM4 market share, on a combined basis across NVIDIA, Google, AMD and others, at the mid 50% range for SK hynix, the high 20% range for Samsung Electronics, and the high 10% range for Micron. The picture is one in which SK hynix holds its volume advantage on the strength of mass production stability, while Samsung Electronics seeks to expand share through a technical spec advantage.
Against this backdrop, analysts say SK hynix's push to speed up the 1c transition will translate into tangible benefits in cost and productivity, beyond simply shrinking the node. Since bit output per wafer rises compared with the previous generation, more bits are produced from the same wafer input, improving cost competitiveness, which is cited as a factor that will help defend DRAM segment profitability from the second half onward. With the 1c transition proceeding alongside a growing mix of high value added products for servers and HBM, analysts say productivity gains are highly likely to feed directly into margin improvement.
An official in the semiconductor industry said, "The pace of the 1c transition itself is encouraging, but it only becomes meaningful if actual production yields and customer qualification schedules back it up," adding, "Starting with HBM4E, the fine process competition between Samsung Electronics and SK hynix will intensify further."
0·-Neutral
m
meme9/6meme
quote: IA's analysis is always excellent. Just to add a few lines to his comment on memory:
The reduction in HBM stack height is partly because the need for it has diminished, but there is clearly another side to it. As high bandwidth HBM was demanded, meeting the bar for high bandwidth HBM was eating up DRAM wafers at too high a rate, so stack heights were lowered to make supply more elastic.
Does that mean HBM itself has become less scarce? I don't think so. Instead of stacking higher, I think a different dimension of scarcity will be emphasized: HBM's high pin speed. In other words, wafer consumption and supply capacity are now being consumed by pin speed rather than by stacking.
Put differently, you could also read it this way: each individual HBM layer has become so precious that the opportunity cost of stacking them up and failing has grown too large.
The HBM spec downgrade is real. Feynman, the generation after Rubin Ultra, was also lowered to 8-High, wasn't it? | Hot Chips 2026: Irrational Recap https://irrationalanalysis.substack.com/p/hot-chips-2026-irrational-recap?r=28k8q1&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true
0·-Neutral
m
meme9/6meme
quote: IA’s analysis is always excellent. Just a few additions to his comments on memory:
The HBM layer count was nerfed partly because fewer layers are now needed, but there was clearly another consideration: meeting the specs for high-bandwidth HBM was consuming too many DRAM wafers, so reducing the stack height was also a way to improve supply elasticity.
That does not mean HBM itself has become less scarce. Instead of stacking more layers, I think the scarcity premium will increasingly shift toward another dimension: achieving higher HBM pin speeds. In other words, bits and supply capacity will now be consumed by pin speed rather than stack height.
Put differently, each individual HBM layer has become so valuable that the opportunity cost of losing it to a stacking failure has grown too high.
HBM de-spec is real. As far as I know, Feynman, the generation after Rubin Ultra, has also been reduced to 8-high. | Hot Chips 2026: Irrational Recap https://irrationalanalysis.substack.com/p/hot-chips-2026-irrational-recap?r=28k8q1&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true