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.

Korea, Eugene Investment & Securities: On the NVIDIA Rubin Ultra HBM despec NVIDIA appears to be discussing a plan to configure the main version of Rubin Ultra with a 2 die structure, HBM4 8Hi, and 8 stacks, and to add an HBM4E 8Hi version at a later point. Assuming NVIDIA's 2027 CoWoS capacity allocation is 1,200k, NVIDIA's accelerator output is estimated at 9.9M units and Rubin Ultra output at 1.6M units. If Rubin Ultra shifts from HBM4E 12Hi 384GB to HBM4 8Hi 192GB, NVIDIA's 2027 HBM demand would be revised down by roughly 10%, from 24.1bn Gb to 21.6bn Gb, and total HBM demand could decline by about 4%. Of course, this assumption does not reflect the possibility of a further expansion in accelerator output. That said, we judge this despec to be a strategy for producing more accelerators out of a limited pool of HBM bits, as the increase in HBM supply from DRAM makers has failed to keep pace with the speed of TSMC's front end and back end capacity expansion. DRAM makers find it difficult to expand HBM capacity flexibly because of the extreme shortage in commodity DRAM. With consumer set makers in smartphones and PCs experiencing production disruptions due to memory shortages, it is also not easy for memory makers to cut allocations to these customers any further.

SITUATION EXPLAINED: The Treasury Secretary gave the frontier labs a D-minus on community outreach. "Whether it's the data center companies, the hyperscalers, or the large language labs like Anthropic, OpenAI, they have done a horrendous job reaching out to communities. I don't blame people for having a backlash." "I used to be a professor and I was known as a tough grader, but it's easy to give these guys a D minus. They have taken their credibility to the edge." "We need to acknowledge there is a lot of Chinese agitprop in there. These protesters who show up in rural areas, that was not a spontaneous outpouring of the community. It's very well organized." "There is no day after tomorrow that China wins against us. If they were to pull away from us on AI, then nothing would matter." "The Chinese want pause. Even the North Koreans, they want pause." @theojaffee: "They're still framing opposition to AI in terms of opposition to data centers, which is the most common way the general public is opposed to AI right now. But there are other reasons to be skeptical of the continued rapid development of AI that have nothing at all to do with data centers, and that are far beyond these."

DP is confirming there will be no further crypto tax delay, this is the clearest signal yet that January 2027 implementation is actually happening, not just the standard "we'll proceed as scheduled" line from earlier this summer. The reasoning: this has already been delayed three times, and delaying again on top of that would undermine the basic tax principle of "income gets taxed," a line DP officials are explicitly using. The government's newly released tax reform package also doesn't include any provision for further delay, confirming this at the policy-document level, not just in verbal statements. The problem critics are pointing to is sequencing. The Digital Asset Basic Act, which is supposed to establish the actual market rules, issuance, distribution, disclosure, still hasn't passed, with roughly 10 competing bills still stuck in committee. So Korea is on track to start taxing crypto income before the law governing the market itself exists. One industry source flagged the specific mechanical problems this creates: no loss offsetting across gains and losses, and easy avoidance paths through DeFi, meaning the tax burden falls disproportionately on ordinary users trading in won on domestic exchanges, likely accelerating capital flight to offshore platforms. The stock tax equity argument resurfaces too, since the financial investment income tax was scrapped, stock investors pay essentially nothing on capital gains, while crypto investors face a flat 22%. That's the exact fairness gap PPP has used against this tax from the start, and it's likely to reignite debate over reinstating the financial investment income tax as well. DP's response to all of this is to lean harder into finishing the Basic Act this session, want both the legal framework and the tax justification secured within the same legislative session. But the two things that killed every previous attempt, the ownership cap and the bank-majority stablecoin issuer structure, are still unresolved between government, ruling party, and opposition, so there's real doubt whether the Basic Act clears in time regardless of DP's intent. Meanwhile PPP keeps filing delay bills from a different angle, Jung Sung-kook's bill pushing implementation to 2030 being the most recent, positioning tax delay as a tool to use against the government in the upcoming tax reform review process rather than a genuine standalone push.

2026, day 257 Good morning from Asia. ‘Oil Industry Braces for Years-Long Iran War’ (Oil Price); ‘Widening Mideast Instability Squeezes Global Oil Supplies’ (WSJ); ‘Saudi pipeline outage threatens loss of 4% of global oil supply’ (Reuters); ‘Sources within Saudi Arabia are now telling us that the damages to the East-West pipeline will take more than a month to repair due to a lack of spare parts and deeply impacted supply chains’ (Marhelm); ‘Saudi oil facility in Jazan directly hit in the attack’ (Tasnim); ‘Saudi foreign minister warns continued regional escalation risks widening conflict’ (Al Arabiya); ‘Potential Iran-Oman Hormuz deal won’t immediately reopen strait, source tells Iranian media’ (ToI); ‘Saudi pipeline attack exposes potent Iraqi militia threat’ (FT); ‘Houthis' Yemen advance leaves Gulf states with uncomfortable choice’ (Reuters); ‘Iran hardliners said to have launched attacks in defiance of leaders, sabotaging peace deal’ (ToI); ‘Iran requested new Russian Geran drones for attacks on Israeli, American targets amid war’ (FT); ‘Trump says US could stay in Iran and keep oil, like Venezuela deal’ (Reuters); ‘Korea, US to hold high-level defense talks amid Hormuz deployment consideration’ (Korea Times); ‘Turkey FM says Israel wants to destabilize Syria as he visits Damascus’ (Al Arabiya); ‘US Air Force pilots lack sufficient sortie experience for combat operations’ (Military Times); ‘Russia says hit Ukraine rail site ‘used to transport military cargo’ from Europe’ (AFP); ‘Russian drone strikes train behind Boris Johnson’s’ (Telegraph) - “Former prime minister unharmed after attack on railway line between Ukraine and Poland”; ‘Trump says Zelenskyy ‘has to stop’ attacking Russian refineries’ (AFP); ‘Ukraine's race for its own ballistic missile’ (Kyiv Independent); ‘Trump reiterates support for united Ireland, says won't talk about Scotland 'yet'’ (Reuters); ‘National Security Agency launches historic restructuring’ (WaPo); ‘Silicon Valley’s attempt to slow AI race triggers surge of suspicion’ (Telegraph) - “Tech giants told to ‘stop pretending the motivation to slow down is purely altruistic’”; ‘Trump Sees Race Against China as Key Issue on AI Regulation’ (WSJ); ‘China’s intelligence chief warns of risks from AI as ‘new arena for strategic rivalry’’ (SCMP); ‘Brics summit ends with Xi’s AI offer, Modi’s warning on critical minerals as weapons’ (SCMP); ‘For China, H-6N bomber enhances nuclear deterrence and second-strike capability’ (SCMP); ‘Japan has few cards to play one month after Putin’s isle visit’ (Japan Times); In geoeconomics, ‘Dutch start-up completes funding round for encrypted silicon technology’ (FT); ‘Japan's future US investments to focus on power sector: trade minister’ (Nikkei Asia); ‘EU Parliament ready to take tougher line on China, says lead MEP’ (Politico); ‘Trump says he is removing U.S. tariffs on Irish whiskey’ (Reuters); ‘EU shelves entry-exit system in France and Portugal’ (Telegraph); In politics, ‘Marine Le Pen vows to put French citizens first in campaign launch’ (FT): ‘Sweden’s election too close to call after late right-wing surge’ (Politico); ‘Rayner accused of ‘two-tier’ donation rules to protect union funding’ (Telegraph); ‘Reform 'very confident' £72m donations were legal - but Rayner warns of 'backdated' action’ (Sky); ‘Trump calls $5,000 payouts 'easy' to fit into federal budget’ (Reuters); In markets, ‘BOJ and Fed face critical test as market pressure rises on rates’ (Nikkei Asia); ‘Central bankers slam 'dangerous’ Mélenchon’ (Euractiv); ‘Oil Jumps as Shutdown of Saudi Pipeline Deepens Energy Crisis’ (BBG); ‘A 5% Treasury Yield Is Raising New Risks for Markets, Economy’ (BBG); ‘Anthropic has chosen the Nasdaq for its IPO listing’ (Business Insider); ‘China’s Tax Clampdown Raises Fears of Founder Share Disposals’ (BBG)

Some thoughts on the AI trade. Going forwards, I think it'll be mega important to be more selective when picking AI stocks. I think the era of mining for new bottlenecks like gold in a riverbed is exhausted. Basically... I think taking a longer-term, buy-and-hold view to stock picking will become more important now. Plainly - high quality companies. (I explain what this might look like below). In H1, we saw any and every AI beneficary get re-rated from low -> fair -> high valuations. That includes tiny microcaps from Europe and Asia where huge returns were enjoyed if you capitalized on the "bottleneck" trade. However, if we use KOSPI as a proxy for retail froth, it's pretty clear that investors are now positioning way more cautiously than H1. Retail net equity buying has dropped by ~90% since July, and investors are instead depositing cash into bank accounts - deposits at Korea's top 5 banks increased in Jul-Aug to above KRW 1,000T for the first time. This could change quickly once the macro environment clears up with time, but ultimately, I still believe that the core bottlenecks will continue to persist - memory, power, photonics etc etc. We all know the list by now. But...within those bottlenecks...which companies will generate the strongest ROI and cash conversion? I think this is the question that most investors should be asking more intensely given how intricate semiconductor financing has become AND the sheer scale of capital in play. In that sense, which companies have the most/longest visibility over demand into the future? And of those, which companies can hike prices so they don't lose out on any of that demand to competitors? An easy example that comes to mind is NVIDIA. Regardless, I think it's that level of certainty that investors are now looking for, rather than H1 where supply chain mapping carried those outsized returns in small caps. Personally, I'm still expecting momentum in the AI trade to continue and still hold some small cap names from H1 as moonshot bets (also some cash for additional fun bets). But, I've repositioned to have a significantly higher concentration in names I view as the bottleneck winner(s) vs. H1. But TLDR: I think it makes sense to now take a longer-term view to your positioning in the AI trade. Which companies are FCF positive/growing? Which have pricing power and possess critical components/materials/infra that can't be designed/competed away? Going forwards, I think these questions will be more and more important as the AI trade matures.

Muslim woman travels to South Korea and is very angry that Korean food has ‘too much pork’: “Koreans put pork in everything: mayo, ketchup, chips. I had to inspect everything before I ate. South Korea claims to be this ultra-modern country, but refuses to respect 2 billion Muslims!” There’s an easy solution to her problem: don’t go there!

🎉 Google Play Games Deals Week on PC Enjoy 20% off BrownDust2 purchases + a 10x Google Play Points booster! Play the Full-Burst RPG BrownDust2 on Google Play Games on PC and get: 💰 20% off all purchases (up to USD 50) ⭐ 10x Google Play Points 📅 Period: Until August 6, 2026 (Thu), 23:59 (local time) 🌏 Regions: Korea, the U.S., Japan, Europe, and more Joining is super easy! 1. Open Google Play Games on PC and tap "Get Started" on the BrownDust2 Page 👉 2. Click the 20% Payment Discount Promotion banner and save the offer Claim your discount now and don’t miss out! #BrownDust2# #GooglePlayGames#