AZ Sestante Chief Investment Officer Andrea Ciaccio reviews markets for January 2025 — the month a little-known Chinese start-up, DeepSeek, wiped US$593 billion off Nvidia in a single session and forced Wall Street to question the cost of the AI build-out. He explains how DeepSeek’s R1 model achieved its breakthrough, why the sell-off showed unusually broad market participation, and why he remains overweight international equities despite the “Magnificent Seven” sell-off. Read the full commentary below, or download it as a PDF.
Download the full commentary (PDF)
By Andrea Ciaccio, Chief Investment Officer, AZ Sestante — as at 31 January 2025
Market Review
On January 20th, the same day Donald Trump was sworn into office, DeepSeek released R1, a reasoning model that performed on par or better than OpenAI’s o1 in multiple accuracy tests for complex problem solving, mathematics, and coding assistance. The advance did not come out of the blue; in fact, the obscure, 20-month-old Chinese AI start-up had launched its V2 model back in May 2024 and its V3 model at the end of December, with the latter having already demonstrated its capacity to outperform Meta’s Llama 3.1, OpenAI's GPT-4o, and Anthropic's Claude Sonnet 3.5. However, the company had been largely flying under the radar over that period, with only specialised publications and sites actively discussing its innovations. Things went differently this time. It took a week for the buzz generated by the release of R1 to finally explode into the mainstream media and, most crucially, to reverberate through financial markets. On January 27th, the S&P 500 and the Nasdaq 100 fell -1.46% and -2.97% respectively, dragged lower by a -16.97% crash in Nvidia. According to Dow Jones Market Data, the chip giant saw 592.7 billion USD of its market capitalisation evaporate in one day, the largest daily decline by any listed company. Following the latest monster drop, Nvidia now holds the (unenviable) distinction of having experienced 8 of the 10 biggest single-day stock declines in history, with all of them having occurred in the space of just 9 months. Even more interestingly, while the technology sector plummeted -5.58% on the news of the DeepSeek disruption, other sectors such as consumer staples and healthcare rallied more than 2% for the day, that is, the S&P 500 exhibited positive breadth despite the fall, as the number of stocks rising was larger than the number of stocks declining.
DeepSeek grew out of High-Flyer, a quantitative hedge fund founded in 2015 by Liang Wenfeng and two of his classmates from Zhejiang University. The trio started to incorporate machine learning into their investment strategies one year later, a prescient move that led their flagship fund to generate outsized returns over the CSI 500, the benchmark for Chinese mid and small cap companies. High-Flyer grew its assets to a peak of roughly 12 billion USD before suffering steep losses in December 2021 on the back of ill-timed trades at a time of large market swings. Eventually, High-Flyer took two crucial decisions: it opted to focus solely on long-only strategies, thus reducing its assets under management by a third, and it spun out DeepSeek in May 2023. The AI start-up continued to draw from the R&D budget of the hedge fund for its funding and for its computing resources, finally achieving two significant breakthroughs. Firstly, V2 introduced the “Mixture of Experts” (MoE) architecture that splits the model into multiple “experts” and only activates the most relevant ones for each task. Its subsequent version, the V3, features 671 billion parameters though it operates with approximately 37 billion parameters at once, optimizing both performance and resource usage. In contrast, ChatGPT uses the more consistent, but potentially less efficient, “transformer” model, which is like having all experts working on every single task. Secondly, R1 removed human feedback from the reinforcement learning required to enhance reasoning capabilities; instead, it encouraged the model to try several different answers at a time rather than explicitly teaching it on how to solve a problem; the end result was a model that developed reasoning and chains-of-thought on its own, including what DeepSeek called “Aha Moments”.
According to the technical paper published by the firm, its latest, cutting-edge R1 was built in 2 months using just 2,048 Nvidia H800s, a watered- down version of the high-powered H100, and 5.6 million USD for the final training. To be clear, that is a measurement of the marginal cost and not of the original cost of buying the compute, building a data centre, and hiring technical staff. Nevertheless, DeepSeek maintains to have developed an open-source AI model at a fraction of the cost sustained by the American hyperscalers and notwithstanding the US restrictions on the export of the most advanced semiconductors to China. Those claims have been widely disputed. According to Alexandr Wang, the billionaire CEO of Scale AI, “DeepSeek has 50,000 H100, which they can’t talk about obviously”. Dylan Patel of SemiAnalysis is “confident” that DeepSeek’s “hardware spend is well higher than 500 million USD over the company history”, that is, at least 10 times more than what officially disclosed by the Chinese AI start-up. However, that would still compare well with, for instance, Meta, which operates the equivalent of 600,000 Nvidia H100s. Other analysts point out the fact that DeepSeek optimized both its model design and infrastructure with the specific aim of overcoming the lack of memory bandwidth implied in using H800s instead of H100s. While questions about the DeepSeek’s making remain, the techniques and improvements it introduced are undeniable and likely to be incorporated into larger, more sophisticated and, potentially, more expensive models built for even more functionalities and commercialised by other researchers and developers in the AI community.
International Equities
US equities shrugged off the shift in narrative and rallied to cap the best start to a presidential term since Ronald Reagan. The Dow Jones Industrial (+4.70% in USD terms) outperformed the S&P 500 (+2.70%) and the Nasdaq 100 (+2.22%) in January. The “Magnificent Seven” gained on the back of the stellar returns posted by Meta, Amazon and Alphabet, but underperformed the rest of US large caps, breaking a 4-month winning streak. Tesla consolidated the doubling of the stock price it experienced from its October lows, while Nvidia, Apple and Microsoft ended the month in the red. Investors rotated away from semiconductors, the “pick-and-shovels” of the AI gold rush, and piled into software companies, the “new AI beneficiaries”, propelling the Dow Jones Internet Composite Index to its highest monthly close since August 2021. All sectors except for technology rose in January, with financials, communication services, materials and industrials scoring increases of +5% and above. After having trailed the S&P 500 for the second year in a row in 2024, healthcare was the best performing sector, rebounding from oversold levels last seen in 1991.
US equities modestly underperformed the rest of the world (as exemplified by the MSCI AC World Index ex USA TR Index) in January. Europe started the year with a bang, as the DAX and the FTSE 100, Germany’s and UK’s blue-chip benchmarks, broke to new all-time highs. Global investors took advantage of cheap valuations on expectations that the Old Continent will ramp up its spending on infrastructure and defence. All other regions underperformed the general index. It was a rollercoaster January for Japanese stocks, which exhibited weakness in the first half of the month, particularly large-cap exporters, before recovering in the second half. China was a mixed bag, with the CSI Overseas China Internet Index, which tracks the performance of the largest Chinese Big Tech firms, rising and the CSI 300 Index, which tracks the performance of the largest stocks traded in Shanghai and in Shenzhen, falling. Within Asia, South Korea, Taiwan and Singapore posted gains, while India declined on growth concerns. Latin America shone in January, as Brazil exploded higher on the back of its currency recovering from a historic sell-off. All in all, the MSCI AC World Daily TR was up +3.36% in USD terms and +2.19% in AUD terms.
Australian Equities
Australian equities interrupted a 3-month losing streak vis-à-vis their international peers in December, as the MSCI Daily TR Net Australia Index outperformed the MSCI AC World Daily TR Index by the most since September 2024. The S&P/ASX 200 rallied smartly to notch another record- high monthly closure. The S&P/ASX 300 jumped +4.46%%, buoyed by consumer discretionary, financials and industrials, which benefited from increased expectations of the first RBA rate cut coming as early as February. Most sectors posted strong gains ranging from +2% to +4%, although they could not keep up with the general index. Consumer staples lagged in the context of a generally “risk-on” tone across markets, which rewarded cyclicals to the detriment of defensives. Utilities bucked the trend, as Origin Energy fell after downgrading the full year production guidance from its Australia Pacific LNG project. Mid-caps and smaller companies outperformed the Top 20. Finally, growth stocks outperformed value stocks for the 3rd consecutive month.
International Fixed Income
Global central banks started 2025 heading in different directions as economic paths diverge around the world. On January 22nd, the Norges Bank, Norway’s central bank, held rates steady at 4.5%. On January 24th, the Bank of Japan (BOJ) lifted its benchmark rate to the highest level since October 2008 at 0.50%. It is the third hike since the central bank started to normalise its monetary policy back in March 2024. In its post-meeting decision document, it revised up its inflation forecasts, underlining its confidence that wages will continue to rise, and Governor Kuroda indicated that further rate hikes are possible later in the year. On January 29th, the Riksbank, Sweden’s central bank, cut its key policy rate by 25 bps to 2.25%, its lowest level since September 2022, and signalled that its easing campaign may be finished for now as officials assess the impact of their actions so far. On the same day, the Bank of Canada (BOC) lowered its benchmark overnight rate by 25 bps to 3.00%, its sixth consecutive reduction since June. The central bank trimmed the country's economic growth outlook on the looming threat of a 25% tariff on Canadian exports. Finally, the FED left its target rate unchanged at 4.25%-4.50%. During the press conference, Chairman Powell made it clear that the central bank is not in a rush to lower borrowing costs, suggesting that it may remain on hold for some time. He maintained that the monetary policy is “restrictive” but observed that financial conditions are “still somewhat accommodative” and that asset prices appear “elevated by many metrics right now”. One day later, the European Central Bank (ECB) lowered its policy interest rate for the fifth time, from 3.00% to 2.75%, as data from Eurostat indicated that the eurozone economy unexpectedly stagnated in Q4 2024.
The Bloomberg Barclays Global Aggregate Index hedged back to AUD added +0.38% for the month. Yields were lower in the US and in the UK, but higher in continental Europe, while credit spreads continued to trade at historically tight levels.
Australian Fixed Income
Domestic headline inflation came in at +0.2% QoQ and +2.4% YoY for the December quarter, missing expectations of +0.3% QoQ and +2.5% YoY. More noteworthy, the CPI Trimmed Mean, the measure preferred by the central bank, surprised consensus on the downside as it decelerated to +3.2% YoY, its lowest level since December 2021. In addition, retail sales advanced +0.8% MoM in November, trailing forecasts for a +1% gain, despite pre-Christmas and Black Friday discounts. As a result, at the end of the month the cash futures were pricing in a cut in February with a 95% probability. The Australian yield curve steepened further, with the spread between the 10-year yield and the 2-year yield reaching its largest differential since December 2022, as the former rose 7 bps to 4.43% and the latter fell 6 bps to 3.80%. The Bloomberg AusBond Composite 0+ Yr (+0.19%) was modestly higher for the month. The Australian Dollar caught a bid vis-à-vis most developed and emerging currencies, while continuing to soften against the Japanese Yen (JPY).
Real Assets
Global property gained +2.15% in USD terms and +0.99% in AUD terms in December. Australia rebounded strongly to top the list. Europe and Asia outperformed the general index, while weakness in data centres weighed on the US.
Global infrastructure added +2.68% in USD terms and +1.52% in AUD terms for the month. Transportation stocks were stronger across the board, with airports, roads and railroads rising in unison. Renewables underperformed on the back of President Trump’s revived efforts to move the US away from greenhouse gas regulation. Electric utilities were flat or slightly negative on concerns about the prospects for expenditures related to AI.
Alternatives
Returns for Alternatives (+1.00%) were mostly positive across strategic mandates in January, with nearly 80% of hedge funds ending the month in the black. Long/Short equities extended their 2024 leadership starting the year on a high note.
Market Outlook
The prevailing consensus among Wall Street analysts is that US Large Caps are to be preferred over other developed and emerging markets in 2025. The most crowded trades among global investors are long positions in the “Magnificent Seven” and in the US Dollar. Therefore, it must have come as a surprise to many that the MSCI AC World Index ex USA TR Index has been outperforming the MSCI USA Net TR Index since the beginning of the year. European and Chinese equities have been on a tear over the same period, generating returns that, at the time of writing of this note (February 20th) are, respectively, almost double and triple those recorded by the US benchmarks. Moreover, after spiking to its highest level since November 2022 in mid-January, the Dollar Index (DXY) has started to soften, prompting speculation among strategists that it may have peaked for this cycle. Finally, US yields have stabilised despite the cash futures having priced in a deferral of the next FED cut to sometime in the third quarter. As pointed out by an astute commentator, since his inauguration, Trump has been making “America last” when it comes to investment performance. We anticipate that trend to persist in the following months. The primary catalysts for that divergence appear to be the heightened trade policy volatility, marked by recurring tariff threats, impositions and retractions, and the disruption in the technology sector driven by China’s rapid deployment of competitive AI applications, which has now also seen the emergence of a European competitor.
Currently, there are two debates raging on the issue of tariffs. The first centres on whether they are merely a negotiation tool, or a revenue offset to extend and potentially expand the Tax Cuts and Jobs Act (TCJA) that is due to expire in 2025. We lean towards the latter view. During his campaign, Trump has repeatedly floated the idea of a “fiscal swap”, which involves eliminating the federal income tax and replacing it with revenues from tariffs. While he may not be able to achieve that fully, it appears he is serious about ensuring that those “that make money off us with trade” pay “their fair share”. The second discussion concerns the impact on price levels. There is a school of thought that looks back at the trade war of 2017-2018 and extrapolates a similar, non-inflationary outcome on the back of counterbalancing adjustments in the value of global currencies vis-à-vis the greenback. Other analysts are less sanguine. They contend that since 2021 we have lived in an “inflationary” world, that is, in an environment in which central banks have struggled to bring CPI down to target, unlike the pre-COVID era when they were unable to raise it to 2%. They argue that the perception of businesses and consumers may have changed, and that tariffs may trigger a “de-anchoring” of inflation expectations this time. In addition, Treasury Secretary Scott Bessent and the Chairman of the Council of Economic Advisers Stephen Miran have repeatedly expressed their preference for a “gradual” implementation of tariffs. In his research note titled “A user’s guide to restructuring the global trading system”, Miran outlines a hypothetical scenario where “the US can proceed to gradually implement tariffs […]. It might announce a schedule, for instance, a 2% monthly increase in tariffs on China, in perpetuity, until the demands are met”. In short, “gradualism” may translate into persistent price pressures rather than a one-time shock.
The news cycle has shortened significantly since the advent of the new administration. Things can now change in a matter of weeks, or even days. The jury's still out, but, that said, we do not expect tariffs to be inflationary, at least not in the first half of 2025. Tariffs and/or the threat of tariffs create uncertainty and hold the potential to disrupt an already late cycle US economy, which has been losing momentum recently, rendering it more vulnerable to shocks. For that reason, we continue to think that the risks to economic growth outweigh the risk of an inflation revival. We see an increased probability of the macroeconomic backdrop shifting from one of a re-acceleration of the economy ignited by “animal spirits” to a “growth scare”, which could provide coverage for the FED to cut more and sooner than what is currently anticipated. If our thesis is correct, in 2025 the central bank may end up having to cut by the 100 bps it was projecting at the end of September and resume its easing campaign sometime in the second quarter. Notably, the minutes from the FOMC meeting held at the end of January reveals concerns about “potential swings in reserves over coming months related to debt ceiling dynamics”. As a result, “various participants noted that it may be appropriate to consider pausing or slowing balance sheet runoff until the resolution of this event”. We suspect that a "temporary" suspension of the Quantitative Tightening (QT) program could naturally lead to its premature end, with positive ramifications for long term yields. Hence, we remain comfortable maintaining a fixed income exposure characterised by a duration comparable to that of the benchmark, particularly in Australia, where we think that rates are too high relative to economic reality.
Let's turn now to the sudden disruption caused by the emergence of Chinese, and now European, artificial intelligence. In fact, DeepSeek is not the only player in this space. Other major Chinese companies, such as Alibaba and ByteDance (the parent company of TikTok), have also released their own AI applications. Moreover, a French company named Mistral has launched an AI app that has been trained and optimised using the NVIDIA’s hardware and software ecosystem, further demonstrating the global proliferation of cost-effective AI solutions. While we are not experts in artificial intelligence, in our opinion the widespread belief that the future of AI will require billions of dollars of capex and ever-increasing amounts of power and energy to develop appears to have come into question, and that may pose a threat to the American “exceptionalism”. A month ago, the consensus was that AI was the domain of a few US hyperscalers with deep pockets. Today, we see a more diverse and competitive global market for AI, with innovative, high-performing but cheaper models emerging from various regions. This shift challenges the prevailing investment thesis that favoured the "Magnificent Seven" tech giants. The implications for NVIDIA and the semiconductor industry are complex. While the initial reaction was a sell-off due to fears of reduced chip demand, the long-term impact is less clear. The proliferation of AI could lead to increased demand for computational power as more companies and individuals adopt AI technologies. This could offset any potential reduction in chip demand from more efficient AI models.
The “Magnificent Seven” are extraordinary companies that, owing to their capacity to generate superior earnings growth consistently, have evolved into huge monopolies, becoming de-facto “cash machines”. However, their dominance may moderate going forward as their aggregate net income growth, while still in double digit territory, is poised to decelerate in 2025 and in 2026. At the same time, figures for the rest of the S&P 500 are projected to sequentially increase, narrowing the gap between the two groups. In addition, the “Magnificent Seven” have recently announced the astonishing capital expenditure numbers required to build out, among other things, AI infrastructure, although it remains unclear whether those investments will convert into higher revenues, margin expansion and profit growth. Meta, Microsoft, Amazon, and Alphabet alone will invest a cumulative 325 billion USD this year, an increase of +46% YoY. The plan is for this spending to pay off in the long run, but there exists a not insignificant risk that it may instead come at the expense of shareholder return. Putting it all together, we will maintain an underweight exposure in the “Magnificent Seven”, primarily due to the concentration risk, while remaining overweight in US Large Caps. We have long favoured a more diversified portfolio, and we think that a “broadening” of the rally leading to a more “balanced” market may be in the cards. While these stocks have performed exceptionally well in recent years, we believe that investors may begin to reposition as new opportunities emerge. We remain overweight in international equities and are comfortable with our current exposure, believing that the proliferation of AI will create opportunities beyond the traditional tech giants.
Download this commentary as a PDF
AZ Sestante is a specialist investment consultant focused on designing and managing a range of multi-manager model portfolios via SMAs, MDAs, and fund of funds. Our parent company Azimut is Italy’s largest independent asset manager listed on the Italian stock exchange. The group manages over AU$55 billion in assets globally including over AU$6 billion in multi-manager solutions. E: [email protected] | www.azsestante.com
Important information
The Morningstar Historical Corporate Sustainability Score is a weighted average of the trailing 12 months of Morningstar Portfolio Corporate Sustainability Scores. Historical portfolio scores are not equal-weighted; rather, more-recent portfolios are weighted more heavily than older portfolios. Combining the trailing 12 months of portfolio scores adds consistency while still reflecting portfolio managers' current decisions by weighting the most recent portfolio scores more heavily.
ESG pillar scores are displayed as a number between 0 and 100 with most scores range between 0 and 25. It is the asset-weighted average of the company environmental, social, governance risk scores for the covered corporate holdings in a portfolio. The scores measure the degree to which a company's economic value may be at risk driven by environmental, social, and governance factors. The risk represents the unmanaged risk exposure after taking into account a company's management of such risks.
This document has been prepared by AZ Sestante Limited, ABN 94 106 888 662, AFSL 284 442 (AZ Sestante). This document is not an offer of securities or financial products, nor is it financial product advice. As this document has been prepared without taking account of any investors' particular objectives, financial situation or needs, you should consider its appropriateness having regard to your objectives, financial situation and needs before taking any action. Past performance is not a reliable indicator of future results. Although specific information has been prepared from sources believed to be reliable, we offer no guarantees as to its accuracy or completeness. The information stated, opinions expressed and estimates given constitute best judgement at the time of publication and are subject to change without notice. Consequently, although this document is provided in good faith, it is not intended to create any legal liability on the part of any other entity and does not vary the terms of a relevant disclosure statement. All dollars are Australian unless otherwise specified.