PremiumTop Goldman Trader Warns The Market Is "Increasingly Short Time"; Here's How To Trade It
The market is increasingly short time...
Rate and equity vol are high, CDX HY is widening towards Liberation Day levels, Brent is at ~114, equities are down nearly 10%, ex-US growth forecasts are starting to be downgraded, and the conflict is now a month old.
...and Goldman Sachs global custom equity baskets team, led by Louis Miller, believe low quality in its various manifestations is on the precipice of breaking lower after spending most of the past month chopping around due to positioning dynamics as well as ever-changing views on the durability of this geopolitical conflict.
Miller outlines three major themes heading into a holiday-shortened week, month/quarter-end, and more headline violence...
1. Stagflation
Stagflation concerns have been dominating the market as oil prices remain elevated.
One month into this conflict and an end is unclear.
We are unwinding the growth expectations we had at the beginning of the year and are seeing multiple scenarios where inflation can spike.
We like buying our Stagflation Pair (GSPUSTAG) in this macro environment to hedge against rising possibilities around “stagflation risk”. Both baskets are liquid to trade $400mm per day with no name exceeding 10% of its ADV. GSXUSTGL consists of commodity equities (oil & gas, gold, industrial metals, agriculture and commodities CAPEX, and defensive compounders like utilities, pharmaceuticals, and high quality staples) that screen favorably during times of stagflation. GSXUSTGS on the other hand is composed of low quality consumer distortionary, semis and hardware, consumer finance and regional banks, cyclicals that consumer oils, as well as unprofitable tech-like companies that screen unfavorably during stagflationary environment.
The pair silently outperformed ~12.5% from year-to-date lows and we expect this trend to continue:
Source: Goldman Sachs FICC & Equities, Bloomberg, as of March 2026. Past performance is not indicative of future results.
A call spread collar selling the 87% put allows you to buy 108% 120% cal spread on GSTMTDAT costless (54 delta, 55v, 48v, 39v respectively, max loss unlimited).
Source: Goldman Sachs FICC & Equities, Bloomberg, as of March 2026. Past performance is not indicative of future results.
Bond Vol (MOVE Index) vs stagflation pair trade (GSPUSTAG Index)..
Source: Goldman Sachs FICC & Equities, Bloomberg, as of March 2026. Past performance is not indicative of future results.
Our commodities strategists now expect oil prices to average above $100 in March and reach $115 in April before falling to $80 at the end of the year, reflecting their expectation that flows of oil through the Strait of Hormuz will remain very low for 6 weeks. They estimate that Brent would peak at $140 in an adverse scenario where flows are disrupted for 10 weeks and at $160 in a severely adverse scenario where flows are disrupted for 10 weeks and infrastructure damage causes a persistent hit to production.
Source: GS Global Investment Research, The Risks to US Inflation from the War with Iran, as of 23-March-2026.
GS economist explain most of the impact of the war on US inflation will come from higher oil prices.
Our rule of thumb is that a 10% increase in oil prices raises headline PCE inflation by 0.2pp and core inflation by 0.04pp.
Much of the impact of higher oil prices on core inflation comes through transportation services—we estimate that a 10% increase in oil prices raises transportation services prices 0.15-0.20pp.
Source: GS Global Investment Research, The Risks to US Inflation from the War with Iran, as of 23-March-2026.
Positioning levels in the pair is at an all-time low, on both the long and short side:
Source: Marquee PlotTool Pro as of 27 Mar 2026, past performance is not indicative of future returns.
In Europe, The market has undergone a sharp transition from “Goldilocks” to pricing stagflation concerns, driven by the surge in commodity prices (GS now sees Brent averaging $110 vs $98 in March/Apr). Financial conditions are tightening on geopolitical uncertainty and market optimism to start the year has cracked, with limited investor appetite to re-gross here. As the path to a clear geopolitical off-ramp remains unclear, we recommend our EU Stagflation Pair (GSPESTAG Index) to hedge lingering inflation and recession tail risks. This strategy leverages high-quality defensives and commodity-geared players against margin vulnerable cyclicals. In the current choppy market environment, we like limited loss hedge implementations.
- GSXESTGL May 103% Call costs 1.3% 30d, GSXESTGS May 95% Put costs 3%, 34d
We also like going long our market neutral US Theme of Themes Portfolio (GSXXMKTP Index), which is +4.38% in March so far with a beta of ~0 and a bias towards energy, metals, AI power, memory, robots, and GSIBs versus shorts in housing, high levered consumer, high yield debt, analog semis, and office REITs for the month of April.
Source: Goldman Sachs FICC & Equities, Bloomberg, as of March 2026. Past performance is not indicative of future results.
2. The Next AI Trade - Bottlenecks
Over the past year, capital has rotated from high-valuation US tech mega-caps towards Asia Memory leaders, primarily in South Korea and Taiwan, as investors looked to diversify and add breadth in the AI supply chain.
CEO Huang's projection of at least $1 trillion in AI infra revenue has sparked a search for the next big opportunity. We believe this "next mover" is centered on niche components and specialized materials in Asia which form a critical bottleneck for AI infra development and we have already started to see early signs of inflows into this theme.
Hierarchy of (AI Bottleneck) Needs
Source: Goldman Sachs FICC & Equities, Bloomberg, as of March 2026. Past performance is not indicative of future results.
This shift is underpinned by a structural supercycle driven by explosive demand for AI hardware and supply-demand imbalance where high utilization rates and restricted capacity expansion for critical components—such as MLCCs, ABF substrates, and tantalum powder—are driving price stabilization and growth. For instance, Resonac announced price increases of over 30% for its copper-clad laminates (CCL) and prepregs, Mitsui Kinzoku is proactively implementing price hikes for its electrolytic copper foil and tantalum powder is seeing rapid demand expansion due to its adoption in NVIDIA’s GB300 and ASIC servers.
We like gaining exposure to this theme via our Asia AI Bottlenecks (GSXABOTL) basket which comprises of stocks under key industries that are crucial to AI development such as MLCCs, ABF Substrates, BT Substrates, PCBs/CCLs, Glass Cloth, Copper Foil, Tantalum Powder, InP Substrates, Memory, Equipment & other semiconductor materials (wafers, photoresists).
Source: Goldman Sachs FICC & Equities, Bloomberg, as of March 2026. Past performance is not indicative of future results.
In the US, we have released our Optical Networking basket (GSXUOPTI Index), Liquid Cooling (GSXUCOOL Index), Inference (GSXUINFS Index), and rebalanced our flagship Datacenters Basket (GSTMTDAT Index).
Long AI Data Centers (GSTMTDAT) – We have adjusted our AI data centers basket in order to include the newest categories of this theme: Liquid Cooling, Optical Networking, and Inference beneficiaries will be added to the GSTMTDAT basket. The basket now includes non-residential construction companies involved in building AI data centers, equipment that goes into data centers like semiconductors, cooling equipment, cables and other hardware. Two sub themes here are optical networking and liquid cooling & water infrastructure:
Long Optical Networking (GSXUOPTI) – To move the massive amounts of data required for AI training and inferencing, hyperscalers and datacenters need increased access to optical transceivers capable of rates of 800 gigabits per second (Gbps) or even 1.6 terabits per second (Tbps) and higher. These crucial components could be in short supply in the near term: production of 800-Gbps transceivers is expected to fall 40% to 60% short of demand through 2027, and shortfalls of 30% to 40% in the supply of 1.6-Tbps transceivers are likely to persist through 2029, according to McKinsey.
Long Liquid Cooling & Water Infrastructure (GSXUCOOL) – Established air-cooling systems are struggling to meet the growing demand of AI data centers and newest equipment for high-performance computing, data management, cloud computing, and generative AI technologies. Liquid cooling appears to be a solution to this demand and our desk believes estimates underestimate this demand. This rising demand is straining local grids and water resources. In the US, data centers currently account for less than 0.1 percent of total water use, but consumption is expected to triple by 2030. This growth, however, is not equally distributed, with approximately 40 percent of current US data centers located in high water-stressed areas. In hot spots such as Ashburn, Virginia, data centers are driving water demand and could account for up to 90 percent of local industrial water use by 2030.
Both Liquid Cooling (GSXUCOOL) and Optical Networking (GSXUOPTI) have lagged the AI data center trade (GSTMTDAT):
Source: Goldman Sachs FICC & Equities, Bloomberg, as of March 2026. Past performance is not indicative of future results.
Europe’s AI bottleneck trade is Power (GSXEPOWR) - EU power demand is expected to grow 1.5-2.0% pa through 2029 and then accelerate to 2.5-3.0% by 2030-31 on our Utilities’ analyst’s base case of electrification of mobility (EVs) & buildings (heat pumps) and demand from AI datacentres. But their latest analysis on agentic AI implies potential for power demand to rise to 3.5-4.0% per annum. This rising power consumption is likely to support higher infrastructure spending and higher margins for power producers (renewables, flexgen), driving an earnings super-cycle for our Power Up Europe basket.
Add to that, the current energy shock where Europe is arguably the hardest hit market given its energy import dependence. This has catalyzed a re-emphasis on EU policy to bolster the region’s energy independence with recent commitments from the European Commission to invest €330mn in fusion energy and nuclear technology.
In that context, the 8.25% MTD sell-off in GSXEPOWR appears excessive and is likely a combination of momentum unwinds, a broad de-risking out of European equities, and a negative rates impact to long-duration assets. GSXEPOWR is our preferred implementation for clients looking to play selective de-escalation trades in Europe, with a positive structural earnings angle. GSXEPOWR May 107/117% call spread costs 1.90% [>5x payout].
Source: Goldman Sachs FICC & Equities, Bloomberg, as of March 2026. Past performance is not indicative of future results.
3. Momo No Mo...
Thursday’s selloff in High Beta Momentum (GSPRHIMO, -7.3%) was the largest down day for the factor since the February momentum unwind and is driven by the selloff in the long leg (Past Winners, GSXUHMOM).
As momentum continued to be resilient amidst the macro backdrop (Still +20% YTD), we noted last week that de-grossing is one of the key risks to the momentum trade given the extremely elevated net factor exposure remaining on our PB Book.
Today’s risk-off tape feels like more of an unwind in some of the trades that have already moved a lot including Memory (GSTMTMEM, -6.9%) and Software vs Semis (GSPUSOSE, +5%), which have both become poster children for the momentum trade.
Realized vol of the factor pair continues to be elevated and this is unlikely to change if the market continues to be headline driven.
Given the positioning dynamics, absent any positive shocks, there could be risk that this unwind persists in the short term, particularly into the weekend given Thursday and Friday returns have lagged Mon-Wed returns by over 7% (H/T Matt Kaplan). We continue to like hedging momentum exposure by buying limited loss expressions on the winners
Source: Goldman Sachs FICC & Equities, Bloomberg, as of March 2026. Past performance is not indicative of future results.
GS Prime Book Net Exposure Remains Elevated
Source: Goldman Sachs Prime Brokerage Services, as of March 2026. Past performance is not indicative of future results.
Tl;dr:
Miller and his team see three main trade themes as the conflict transforms from a 'short-term military operation' to a full blown war...
On the long/short RV side, we like the symmetry around our US stagflation pair (GSPUSTAG Index)...
...which looks too low relative to the accelerating stagflation narrative, rising rate vol, and interesting positioning dynamics of both legs (long leg positioning low, short positioning high).
We also like going long our market neutral US Theme of Themes Portfolio (GSXXMKTP Index), which is +4.38% in March so far with a beta of ~0 and a bias towards energy, metals, AI power, memory, robots, and GSIBs versus shorts in housing, high levered consumer, high yield debt, analog semis, and office REITs for the month of April.
We think every weekend and business day without resolution will be increasingly unforgiving and think it’s time to focus on classic high beta risky names.
As such, we like buying the 2 month 75 92.5% put spreads on our Low Quality Themes (GSXULOWQ Index) for 3.3% (max loss premium paid, 21 delta).
We also like taking a second look at our extensive menu of low quality short expressions including: HY Ex Energy (GSXUHYXE Index), Non Profitable Tech (GSX1NPTC Index), Non Secular Non Profitable Tech (GSCBNOPS Index), Unprofitable Small Caps (GSX1LILP Index), and the Most Shorted index (GSX1MSAL Index) on the short side.
On the offensive side, we are hyper focused on AI bottlenecks and the next manifestations of the AI trade and want to buy meaningful dips here.
Our team recommends our Asia Bottlenecks basket (GSXABOTL Index), US Datacenters (GSTMTDAT Index) and EU Power (GSXEPOWR Index). You can sell the 87% put to buy the 108% 120% call spread on GSTMTDAT as costless call spread collar (54 delta, 55v, 48v, 39v respectively, max loss unlimited).
In Asia, AI Bottlenecks (GSXABOTL Index) in industries that are crucial to AI development such as MLCCs, ABF Substrates, BT Substrates, PCBs/CCLs, Glass Cloth, Copper Foil, Tantalum Powder, InP Substrates, Memory, Equipment & other semiconductor materials (wafers, photoresists).
Europe’s AI bottleneck trade is Power (GSXEPOWR) given EU power demand is expected to grow 1.5-2.0% pa through 2029 and then accelerate to 2.5-3.0% by 2030-31.
In the US, we have been monitoring Inference (GSXUINFR), Optical Networking (GSXUOPTI), and Liquid Cooling (GSXUCOOL).
We like pairing these exposures with momentum hedges until greater clarity on this conflict is reached
Finally, two questions that were top of mind inbounds this week from investors...
Why hasn’t Defense delivered?
Geopolitics has been the driving force of markets for the last four weeks and yet Global Defense (GSXGDFNS) has declined -9.26% since the start of the conflict. Korea Defense (GSXAKDEF) is the only regional defense basket that has delivered positive returns (+6.62%) this month.
It feels like the initial valuation re-rate story is done and markets are now acutely focused on near-term fundamentals, where absent an immediate upgrade to earnings we are unlikely to see a significant price bounce.
Longer-term debates around traditional artillery versus advanced defense tech (drones, missiles) have fueled concerns on sustainability of EPS momentum for conventional warfare names. Worth flagging, we have had increased incoming on our Global Drones (GSXGDRON) basket.
What do you think of Google’s TurboQuant announcement (6x less memory, 8x performance boost?)…
While memory names in particular could face near-term volatility due to recalibrated demand expectations, we believe the structural outlook for HBM leaders remains exceptionally robust in the long term. Optimization technologies such as TurboQuant, while reducing aggregate memory requirements, significantly elevate the performance benchmarks for data throughput. As GPU processing capabilities accelerate by up to 8x, the necessity for high-performance memory bandwidth becomes a critical dependency to eliminate computational latency and prevent processor idling. This shift favors industry leaders like SK Hynix and Samsung Electronics, who possess the technical expertise to deliver the specialized, high-speed memory required for next-generation AI architectures.
Hence, as “Jevons Paradox” suggests, making AI faster and cheaper will actually explode total demand, expanding the overall AI Pie, as companies find new ways to integrate AI in their workflow.
And don't forget, the new weekly seasonal bias...
Thu–Fri sessions have been painful for S&P. Over the last 90 trading days, Thu–Fri cumulative returns have lagged Mon–Wed by ~7.15%.
Source: Goldman Sachs FICC & Equities, Bloomberg, as of March 2026. Past performance is not indicative of future results.
Notably, nearly all of this divergence (~7.28%) has occurred since Feb 28, coinciding with the Iran escalation, as investors appear to de-risk into the weekends.
Will this week be 'more of the same'... and of course, this week also brings payrolls - what if it's negative again?
Professional subscribers can read much more from Goldman's Sales & Trading team here at our new Marketdesk.ai portal
