EM Technology
Early summer pullback masks improving fundamentals
26 August 2026
- Technology shares retreated sharply at the start of Q3 2026 following exceptionally strong gains in H1.
- Key causes included: exceptionally strong year-to-date performance, position concentration and elevated leverage build-up, strategic weighting adjustments, concerns on capital raising, capital expenditure levels and AI returns on investment.
- Looking through the near-term volatility, industry fundamentals continued to improve, with increased visibility on AI demand, cloud backlogs and increasing confidence in long-term returns on AI investment.
- We continue to like the exposure to EM tech with positive earnings revisions combined with lower share-prices have left the MSCI Emerging Market (EM) Information Technology Index trading at 25-year valuation lows (on an earnings basis), offsetting the risks around capital expenditure, capital raising and supply-chain cyclicality.
The start of Q3 2026 saw a sharp correction across technology stocks following exceptionally strong gains in H1. During July, the Philadelphia semiconductor index (SOX) fell 21%, while the MSCI Korea Information Technology Index declined 27%.1 Given that the big two memory stocks SK Hynix and Samsung Electronics account for 95% of the Korean information technology index, the sell-off was particular severe, with the index falling 49% from its 19 June peak and 30 July trough. The MSCI Taiwan Information Technology Index was down 14%, although the index-level move understated the weakness in individual stocks given Taiwan Semiconductor Manufacturing Company’s (TSMC) 64% weighting and relatively low beta.2
Chart 1: MSCI EM Information Technology Index
Source: Bloomberg, as at 18 August 2026.
Several factors contributed to this pullback, although some degree of correction should not have come as a big surprise given the magnitude of the prior year rally. Even after July’s decline, YTD returns at the end of the month remained exceptionally strong, with the SOX up 60%, the MSCI Korea Information technology Index up 63% and the MSCI Taiwan Information Technology Index up 61%.3
We see the recent weakness as a healthy correction rather than a signal of a deteriorating outlook.
Positioning, leverage and funding concerns amplified the pullback
One of the key catalysts for the correction was likely position reductions by large strategic investors at quarter-end as they rebalanced portfolios back to target weights and risk limits. Given the scale of YTD gains, these adjustments may have generated significant selling pressure. Elevated retail participation and leverage further amplified volatility, with margin calls exacerbating the initial sell-off. This was particular evident in Korea, where both single-stock and index-based leveraged ETFs (exchanged traded funds) experienced strong inflows.
While high retail participation and the use of leveraged products have been a broader regional phenomenon, Korea attracted the greatest attention following the launch of several leveraged single-stock ETFs, some offering as much as 10x exposure, on 27 May 2026. According to a Bloomberg markets report, these products attracted approximately USD 9 billion of inflows within two months. At the peak of the market volatility in June, SK Hynix, Samsung Electronics and their associated leveraged ETFs accounted for around 70% of total daily trading activity in the Seoul stock market.
Additional concern centred on the potential impact of large-scale capital raising by both existing companies and forthcoming IPOs (initial public offerings). Several high-profile listings are expected over H2 2026, including potentially OpenAI and Anthropic4, which recently completed funding rounds in the private markets of approximately USD 122 billion5 and USD 65 billion6 respectively, highlighting the substantial financing needs of leading AI developers while IPOs of both OpenAI and Anthropic. A busy pipeline of potential IPOs and follow-on issuance has raised questions about whether equity markets can continue to absorb elevated levels of capital raising without impacting valuations.
In its 4 August 2026 report, SIFMA (Securities Industry and Financial Markets Association) reported that US equity issuance had reached USD 302 billion YTD through July, up 114% year-on-year (YoY)7. Annualising issuance over the January-July period would imply full-year US equity issuance of approximately USD 550-650 billion. A midpoint estimate of USD 600 billion is equivalent to 1.85% of US nominal GDP (versus upper bands of 2.0-2.2% of GDP) and 0.8% of US market capitalisation (elevated peaks levels of 1.6-1.8% in 2021). While elevated, these levels do not appear alarming levels at this point.
Credit issuance form the hyperscalers has also increased, with JP Morgan increasing their 2026 US TMT (Technology, Media and Telecom) corporate credit issuance forecast from USD 450 billion to USD 540 billion, primarily due to increased AI spend. If you annualise the January to July run rate, this implies a full year estimate of total US corporate supply of USD 2.9 trillion, of which AI related issuance is just part of the supply story this year while spreads and credit default swaps levels (CDS) indicate this is being well absorbed.
Despite increased equity and debt issuance to fund AI-related investment, US investment-grade CDS spreads remain near historical lows, suggesting credit markets continue to absorb the additional financing requirements without signs of significant stress.
Chart 2: US investment grade CDS spreads remain near historic lows
Source: Bloomberg, as at 18 August 2026.
The emergence of lower-cost frontier models and declining token prices for closed AI models has also contributed to bouts of volatility. However, we believe these concerns are largely misplaced. Lower costs could ultimately accelerate AI adoption, broaden the range of commercial applications and support long-term demand growth across the AI ecosystem.
While these factors contributed to near-term volatility, developments across the AI ecosystem over the summer suggest the underlying fundamental outlook has continued to strengthen.
Hyperscaler demand reinforces confidence across the AI supply chain
While there are certainly risks to monitor, industry developments over the summer, particularly Q2 results from the leading technology and CSPs (cloud service providers) / hyperscalers were highly encouraging.
TSMC management expressed rising confidence in AI demand, with an even more upbeat tone than the previous quarter. Based on customer and end-customer deployment timelines and growth in product roadmap, management are increasingly positive on both current and future sustainability of demand, with visibility now extending into 2029-30. Recent strong order flow led to another upgrade to full year 2026 revenue growth guidance to over 40%, and the firm also a raised its capital expenditure (CapEx) guidance to meet these opportunities.
In the memory space, Samsung Electronics gave an upbeat assessment of demand/supply for memory into end 2028 and gave more specific guidance on the industry trend of locking in long-term contracts for their future output. Jaejune Kim, Executive Vice-President of Memory Business, stated, “The (memory) supply constraints are expected to become even more severe in 2027 than 2026, reinforcing our view that the supply shortage will persist through 2028” and “once we finalise on the currently pending contracts as well, we believe that DRAM8 and NAND9 supply volumes under multi-year supply contracts will easily account for 60% to 70% of our planned capacity, based on our current mid to long-term production plan”.10
SK Hynix management was less specific on timelines but positive on an extended cycle particularly in leading edge product saying “With increasing demand for high performance compute, the company expects demand for premium products such as HBM (high-bandwidth memory), server DRAM and high-capacity eSSDs to continue to grow for an extended period”11
The increasing confidence in the Asian technology hardware supply chain has been reinforced by strong commentary from the US hyperscalers / CSPs (cloud service providers), including Alphabet (Google), Amazon, Microsoft and Meta.
All reported accelerating cloud sales in Q2 2026. Google Cloud, Microsoft CSPAzure and Amazon Web Services (AWS) grew by 82%, 43% and 37% YoY respectively, significantly ahead of the consensus forecasts12. In addition, cloud backlogs surged in Q2 2026, providing further evidence of robust demand. Amazon was a good example: the stock rose 15% on the trading day after the results announcement13, with AWS backlog up by 36% quarter-on-quarter (QoQ)14.
While you might expect a positive bias from these companies, some of the commentary on high performance compute and AI demand was quite effusive. Amazon’s AWS CEO Matt Garman, said the demand being witnessed by the company for AI-related services is “just massive” and commented “We see large training clusters being used by a number of companies, but as these models get really popular and really powerful, more and more companies are integrating that inference into their workloads”15.
Chart 3: Increasing AI usage among US businesses
The Ramp AI Index measures the share of US businesses with paid AI subscriptions, based on spending data from more than 70,000
firms. Source: RAMP, as at July 2026.
Strong Calendar Year (CY) 2026 guidance and increasingly constructive commentary on CY 2027 led three of the four major US hyperscalers to raise their CY 2026 CapEx expectations. and 2027 guidance. Consensus cloud CapEx forecasts now approach USD 1.2 trillion, representing approximately 30% YoY growth and, according to JP Morgan, around USD 170 billion more than expected to the prior earnings season.
Chart 4: CapEx forecasts for hyperscalers
Source: GAM, Visible Alpha, Morgan Stanley, as at July 2026. Estimates have not been independently verified by GAM. No assurance
can be given that such estimates will prove accurate.
Despite higher CapEx plans, management teams expressed greater confidence in 2026 and 2027 return on capital, supported by rising backlogs, multi-year customer commitments, pricing power and efficiency gains. AWS (Amazon Web Services), Azure (Microsoft’s cloud platform), and Google Cloud all see a sizable runway ahead as enterprise AI adoption remains in its early stages and cloud migration continues.
Importantly, management teams consistently referred to supply constraints, with demand from both external cloud customers and internal AI workloads exceeding current capacity. We believe this suggests the industry remains in the early stages of a broader AI infrastructure build-out rather than approaching a cyclical peak.
As an example, Alphabet CEO Sundar Pichai stated on the results call, “We are working off a disciplined ROIC (Return on Invested Capital) framework here” and “We are seeing strong demand indicators, including long-term deals, the existing deals which we have, which are renewing at exceptional demand on a moving-forward basis. And so we are using all that to plan and invest accordingly.”
Valuation have become increasingly attractive
The combination of strong earnings, upward revisions to earnings forecasts and the recent share-price pullback has left valuations to trading at or below 25-year lows on an earnings basis, while returns on equity have increased significantly16. Although index valuations are influenced by the large weightings of SK Hynix and Samsung Electronics, the recent correction has returned a number of companies with improving earnings outlook to more attractive valuation levels.
Chart 5: EM technology valuations near multi-decade lows
MSCI EM Information Technology Index forward P/E multiples and return on equity
Source: Bloomberg, as at 18 August 2026.
Preferred opportunities within the AI supply chain
Our preferred exposure remains in the upstream segments of the AI supply chain, including foundry, advanced packaging, integrated circuit (IC) design and memory, where oligopolistic industry structures, supply shortage, prudent CapEx outlay and attractive valuations are, in our view, most evident.
The leading Asian memory companies, Samsung Electronics and SK Hynix, remain among our high-conviction opportunities. Their focus on DRAM and HBM (High bandwidth memory) positions them well to benefit from continued growth in AI-related demand. In addition to an oligopolistic industry structure, disciplined CapEx plans, strong free cashflow generation and attractive valuations multiples, both companies are increasingly focused on shareholder returns, with current forecasts implying potential attractive yields.
Memory architecture has become a critical enabler of AI implementation. AI applications require much higher memory specifications than traditional computing workloads, driving demand for advanced memory solutions. A good example is that, AI servers require 8-15 times more memory capacity compared to conventional servers.
Although HBM is the key growth driver, broader memory market dynamics remain supportive. As illustrated in the chart below, DDR517 spot prices, one of the indicators we monitor for the broader DRAM market, continue to show resilient pricing, supporting the view that supply-demand conditions remain favourable across memory markets and according to UBS tight demand/supply conditions are expected to persist until Q2 2028.
Chart 6: DDR5 spot prices - Upward momentum indicative of market tightness
Source: Bloomberg, as at 20 August 2026.
Based on UBS forecasts, SK Hynix is expected to generate cumulative free cashflow equivalent to approximately 68% of its current market capitalisation between 2026 and 202818 while Samsung Electronics is expected to generate around 75%. Combined with higher visibility on shareholder return policies and (as mentioned above) the increasing use of long-term supply agreements, including contracts extending up to five years, this could present an attractive return profile to shareholders.
Samsung Electronics at their recent Q2 results clarified their shareholder return policy which Mirae Securities estimates could lead their 2026 dividend yield to reach 6.8–9.5% for common shares and 9.1-12.7% for preferred shares. SK Hynix has also recently clarified and upgraded their shareholder return policy, where in addition to a KRW 40 trillion share buyback programme in the coming three months, announced that it plans to return more than 50% of cumulative free cashflow in the current policy period 2025-2027E through both buybacks and dividends, with more details to come around Q3 2026 results.
The investment case remains compelling
As highlighted in our previous article, “How Asia is Powering the AI Era”, we continue to believe the investment case for technology remains compelling. Despite the recent pullback, stronger AI demand visibility, accelerating hyperscaler investment and tight supply conditions continue to support the outlook for selected areas of the AI value chain. Combined with increasingly attractive valuations, these developments reinforce our constructive view.