AZ Sestante Chief Investment Officer Andrea Ciaccio reviews markets for September 2025 — a month defined by a frenzy of AI infrastructure deals, from Anthropic’s US$13 billion raise to Nvidia’s US$100 billion commitment to OpenAI, that pushed AI’s share of the S&P 500’s return since 2024 to roughly 72%. He explains why today’s boom differs structurally from the Dot-Com bubble, and why he remains constructive on equities despite mounting “bubble” warnings. 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 30 September 2025
Market Review
The AI trade went into overdrive in September following a frenzy of announcements of mega-deals primarily centred on the infrastructure required to support the technology. Anthropic, the San Francisco-based developer of the family of large language models (LLMs) named Claude, kicked off the month’s funding spree on the 2 nd, by securing a massive Series F round of 13 billion USD. The transaction propelled the firm’s valuation to 183 Bil USD, up from 61.5 Bil USD earlier in the year, and it now stands as the third-largest venture round of 2025, trailing only OpenAI’s record-setting 40 billion USD raise and Meta’s 14.3 billion USD purchase of a 49% stake in data-labelling startup Scale AI. On September 8th, Nebius Group, a European “neo-cloud”, that is, a company that builds and operates data centres using a vast amount of high-performance graphics processing units (GPUs) and then rents those out, announced a 17.4 billion USD, five-year agreement to supply Microsoft with dedicated, advanced capacity. The Amsterdam-based company, which was spun out from Russian internet giant Yandex back in July 2024, rallied +64.33% for the month. Two days later, the Wall Street Journal (WSJ) reported that OpenAI had signed a five-year, 300 billion USD contract with Oracle to secure the computing capacity required to train and operate its next generation of AI models starting from 2027. Oracle shares were up +24.37% in September, adding close to 200 billion USD to its market capitalisation, despite the firm narrowly missing on both the top and bottom line in Q2. On September 18th, Nvidia announced a landmark agreement to invest 5 billion USD in its beleaguered rival Intel. As part of this strategic partnership, the two semiconductor giants plan to jointly develop advanced chips designed for both personal computers and data centres, signalling a new era of collaboration in an industry long defined by fierce competition. Intel shares jumped +37.78% for the month. A few days later, Nvidia signed a letter of intent with OpenAI to provide the latter with as much as 100 billion USD to support the buildout of new data centres and other AI infrastructure. The most valuable company in history, with a staggering market capitalisation exceeding 4.5 trillion USD, will disburse the amount progressively as new gigawatts come online and will take a non-controlling equity stake in the ChatGPT maker as part of the funding. However, much of the investment will, in effect, go right back into Nvidia via hardware leasing or purchasing as OpenAI will deploy at least 10 gigawatts (GW) worth of Nvidia systems. The deal drew attention due to its potentially “circular” financial structure, with the semiconductor behemoth recording sales of its chips financed by money it itself provided, creating a loop that may inflate revenues without necessarily reflecting genuine third-party demand. However, the stock reacted positively to the news, notching its highest monthly close ever. For its part, OpenAI has emerged as the epicentre of a rapidly expanding AI investment ecosystem, now estimated to be worth around 500 billion USD, notwithstanding the fact that the firm doesn’t expect to be cash-flow positive until near the end of the decade. While ChatGPT had 800 million weekly users at the end of September, a significant jump from 500 million just 6 months before, the paying proportion is estimated to be less than 5%. That is not deterring the company from expanding aggressively, with the stated aim of hitting 125 billion USD in revenues in 2029. On September 25th, the world’s largest startup, more valuable than Elon Musk's SpaceX and TikTok's parent company ByteDance, inked a 6.5 billion USD deal with CoreWeave, a “neo-cloud” which went public in March of this year and whose stock has more than tripled since then. The new agreement extends the scope of the partnership between the two organisations, bringing its total value to 22.4 billion USD. Originally founded as a cryptocurrency mining company in 2017, CoreWeave has since reinvented itself as the leading example of a new generation of AI-native cloud providers.
The company’s transformation has been accompanied by explosive growth, from just 16 million USD in revenue in 2022 to an astonishing 3.5 billion USD over the past 12 months. On September 30th, it ended the month on a high note by unveiling the signing of an agreement with Meta to provide the latter with 14.2 billion USD of AI cloud infrastructure. CoreWeave shares were up +32.81% in September. Several analysts have expressed scepticism about the sustainability of the booming AI capital expenditures in the US, which are projected to total 400 billion USD this year and to exceed 500 billion USD in 2026 and in 2027. An astute commentator compared those figures to the paltry 12 billion USD that, according to the WSJ, American consumers currently spend every year on AI-related services. The former is roughly equivalent to the annual GDP of Singapore while the latter to that of Somalia. The contrast between those two economies captures the staggering gap between the lofty ambitions and the current economic reality of inference models. In fact, it remains unclear not only whether companies will ever recoup their massive AI investments, but if they will be able to fulfill their spending commitments. Meanwhile, electricity costs are climbing across the US, and, according to Bloomberg, the increases are especially steep in regions hosting large concentrations of data centres, where wholesale rates have surged by as much as +267% over the past five years. Still, exuberant investors have continued to pile into the theme. From the beginning of 2024 through September 2025, approximately 72% of the S&P 500's return was attributable to AI.
International Equities
US equities extended their upward momentum during the month, climbing to fresh record highs. The S&P 500 (+3.53% in USD terms) recorded its best September in 15 years, while the Nasdaq 100 (+5.40%) logged its 5th best September ever, defying seasonal trends. Returns for the Dow Jones Industrial (+1.87%) were more muted given its marginal exposure to the semiconductor sub-sector, which rose double-digits on the back of the AI spending super-cycle. The Russell 2000 (+2.96%), the popular benchmark for smaller companies, broke out to new all-time highs after a long consolidation. While conventional view holds that investors may be looking to rotate away from high-flying megatech names and into undervalued or out-of-favour small caps, some of the best performing names during the month were found in quantum computing (Ionq, Rigetti Computing and D-Wave Quantum), defence tech (Kratos) and advanced nuclear technology (Oklo), with the latter expected to power AI data centres with small modular reactors. The “Magnificent Seven” trounced the rest of US large caps, with their combined market capitalisation closing in on 21 trillion USD, representing a record 36.4% of the S&P 500. Gains were primarily fuelled by Tesla, as Amazon and Meta were down for the month. In the meantime, new winners have started to emerge in the market. Two of them are Robinhood, the fintech platform focused on commission-free investing and trading, and Applovin, the AI-powered adtech platform, which were added to the S&P 500 on September 22nd following a torrid rally. Technology and communication services reigned supreme in September. Utilities posted strong gains, while all other sectors could not keep up with the benchmark. Materials, consumer staples and energy bucked the trend. Retail investors continued to load up on highly speculative stocks, momentum and high beta, shunning high dividend, value and defensives. US equities performed in line with the rest of the world (as exemplified by the MSCI AC World Index ex USA TR Index) in September. Emerging markets topped the list owing to the vigorous rallies in the AI beneficiaries Taiwan and South Korea and the continuous resurgence of the “Terrific Ten” group of stocks, that is, China’s top tech firms. In Japan, the Nikkei 225 index broke above 45,000 for the first time in history, led higher by exporters and AI-related companies. However, the broader TOPIX index trailed its international peers. Europe lagged as well, hindered by the heavyweight energy and healthcare sectors. All in all, the MSCI AC World Daily TR was up +3.62% in USD terms and +2.48% in AUD terms.
Australian Equities
Australian equities experienced a modest decline in September, diverging from the global risk-on sentiment. The MSCI Australia Net TR underperformed the MSCI AC World Daily TR by the largest amount since December 2024, as the S&P/ASX 300 shed -0.65% for the month. Weakness was broad-based, with nine sectors out of eleven ending September in the red. Materials were the standout performer, propelled higher by miners. Gold and silver jumped +11.55% and +17.03%, recording their best month since August 2011 and July 2020 respectively. Copper rose +7.49% while iron ore prices stabilised following earlier volatility amid expectations that China would expand infrastructure funding. Utilities gained as AGL Energy caught a bid following its double-digits decline in August. Conversely, energy crashed as the WTI crude failed to participate in the commodity rally and the consortium led by XRG P.J.S.C., a subsidiary of Abu Dhabi National Oil Company, which includes Carlyle, withdrew its indicative offer to acquire Santos. Despite the overall market retreat, mid-caps and smaller companies exhibited strong momentum, with the former adding +0.38% and the latter climbing +3.60%. Finally, the Ex-20 index outperformed the Top 20 by the most since February 2024 and growth stocks outperformed value stocks for the second month in a row.
International Fixed Income
Four of the eight major global central banks that met during the month held interest rates steady; those include the European Central Bank (ECB), the Bank of England (BOE), the Bank of Japan (BOJ) and the Swiss National Bank (SNB). On September 6th, the Bank of Canada (BOC) delivered its first cut since March, reducing its benchmark overnight rate by 25 basis points to 2.50%. After pausing in August, the Norges Bank, Norway’s central bank, lowered the policy rate on September 18th, from 4.25% to 4%. This pre-emptive easing came just before the release of the latest inflation data, which painted a mixed picture, with the headline number surpassing market expectations but the core inflation rate coming out marginally lower than forecasts. On the same day, the FED decreased its target rate by 25 basis points, bringing it to a range of 4.25%-4.50%. The vote was 11-1, with one dissent, the newly appointed (and potentially politicised) Stephen Miran advocating for a larger 50 basis points cut. Moreover, the central bank released its summary of economic projections, which upgraded its median forecasts of economic growth, but also of inflation, and trimmed its estimates of unemployment. Despite those overall supportive revisions, the Dot Plot signalled further cuts, reflecting a more accommodative monetary policy stance. Finally, on September 23rd, the Riksbank, Sweden’s central bank, cut its key policy rate by 25 basis points to a three-year low of 1.75%, its third reduction in 2025. The Bloomberg Barclays Global Aggregate Index hedged back to AUD advanced +0.68% for the month as falling long-end rates boosted bond prices. Yield curves were generally flatter across the world, except for Japan, where the 10- year yield rose to its highest level since May 2008.
Australian Fixed Income
On September 30th, the RBA kept the cash rate unchanged at 3.60% and adopted a notably hawkish tone, cautioning that inflation could surprise to the upside in Q3 and signalling a more cautious stance on further policy easing. Its decision was widely anticipated by market participants after GDP grew at its fastest pace in nearly two years in Q2 (+0.6% QoQ and +1.8% YoY), largely driven by higher household spending, the monthly indicator of consumer prices (CPI) accelerated to +3% in August (although the CPI Trimmed Mean ticked down to +2.6%) and the unemployment rate held steady at 4.2%. The data were consistent with the central bank’s observations of a faster pickup in private sector activity as the 75 basis points of reduction delivered between December 2024 and August 2025 filters through the economy and transmits into consumption. At the end of the month, the cash futures were pricing only one additional cut in the back end of H1 2026 to bring the current easing cycle to conclusion. Domestic fixed income remained resilient, with the Bloomberg AusBond Composite 0+ Yr adding +0.10% on the back of tighter credit spreads. The yield curve transposed higher and flattened with the 2-, 5- and 10- year yields 15, 11 and 3 basis points higher at 3.49%, 3.75% and 4.30% respectively. The Australian Dollar appreciated across the board, reclaiming 66 cents vis-à-vis the greenback on a monthly basis.
Real Assets
Global property was up +0.75% in USD terms, but down -0.37% in AUD terms in September. At the regional level, Europe topped the list on the back of a rebound in the UK, followed by the US. Sector performance was mixed during the month, with longer duration segments (healthcare, data centres) outperforming their more economically sensitive peers (self- storage, residential). Global infrastructure added +1.32% in USD terms and +0.20% in AUD terms for the month. Electric and gas utilities and renewables led gains, while communication infrastructure continued to lag. Toll roads were adversely affected by political instability in France, as the government collapsed after Prime Minister Bayrou lost a confidence vote in Parliament.
Alternatives
Alternatives (+1.43%) delivered strong, broad-based positive returns in September 2025. Discretionary macro and trend following mandates led the charge as they recorded their best month since December 2020 and February 2024 respectively.
Market Outlook
The current market cycle has been extraordinary, both in its duration and magnitude. While we have maintained a constructive view on equities even during the depths of the Liberation Day crash, a stance that has been validated by market performance, we must acknowledge that the nature of this rally has been surprising. Investors who have remained fully invested in stocks have enjoyed enormous gains fuelled by enthusiasm for technology in general and for artificial intelligence in particular. As sentiment appears to have entered a euphoric phase, concerns are mounting that the current boom may be taking on bubble-like characteristics. However, market participants are deeply split over this question. According to the latest Bank of America (BofA) fund manager survey, roughly 54% of investors now believe AI-related stocks to have entered bubble territory. The primary evidence lies in the extraordinary concentration of equity market gains in the ten largest AI companies and in the proliferation of newer, unprofitable entrants and “pre-revenue” technology firms with multi- billion-dollar valuations, evoking echoes of the speculative frenzy of the late 1990s. Even the “Magnificent Seven”, the most profitable enterprises in human history, are spending roughly 60% to 70% of their free cash flow on AI development. This is, for all intents and purposes, a massive venture capital investment; it is highly uncertain, and some of these bets will inevitably fail. In a normal environment, investors would demand a lower multiple to compensate for this massive gamble. Today, that risk is being ignored. Sceptics highlight the mismatch between revenues and capital spending and the feedback loops typical of periods when “animal spirits” run wild. The market narrative appears to have become self- reinforcing, as companies fund each other’s expansion, with valuations justifying spending and spending in turn fuelling valuations. As stock prices grind higher and higher, volatility falls and investors take the apparent stability as validation to buy more. Yet, in our view, there is a critical nuance. Unlike previous manias, today’s AI enthusiasm is underwritten by both corporate balance sheets and government policy. The Trump Administration is pursuing a "run it hot" strategy, channelling billions into semiconductor subsidies and national infrastructure. Fiscal spending, regulatory support and political capital are aligned with this technological race. That said, valuations this stretched are inherently fragile. The market’s pricing of the AI transformation may have run too far, too fast, and if the economic cycle softens or policy support wanes, the fallout could be sharp. However, while the sector exhibits symptoms of excess the likes of crowded trades, sky-high multiples and cash- burning companies, the macroeconomic backdrop makes the materialising of a destructive “mean reversion event”, that is, of an abrupt collapse, less likely in the near term. As long as liquidity remains abundant and policymakers are effectively underwriting the sector, the “bubble” can inflate further.
On the surface, similarities between today and the Dot-Com bubble exist, but while the “micro” may rhyme, the “macro” tells a different story. Many of today’s trends feel similar as a handful of AI-driven firms dominate index performance while defensive sectors not only lag badly, but they are actively sold by investors to pile into high-growth technology names. However, the structural context of 2025 is vastly different from that of the Clinton’s era. The then-President embraced a “third way” aimed at synthesising both the “laissez-faire” and the “welfare state” economic approaches. His administration actively boosted the growth of the private sector, but primarily through global trade liberalisation, deregulation and fiscal prudence. His focus was on ensuring American firms accessed global markets, improved productivity and adopted new manufacturing technologies. While he laid the groundwork for the commercialisation of the “World Wide Web”, transitioning it from a government-controlled network to a platform for private innovation, supported advanced manufacturing R&D through investments in education and deliberately avoided heavy regulation of emerging internet businesses, his administration did not pick winners. Today, Washington is actively taking stakes in strategic companies like Intel, framing AI as a national security priority in competition with China, and targeting large-scale domestic industrial resurgence via protection or massive state intervention. Fiscal and industrial policy are explicitly aligned with sustaining tech leadership. Moreover, Clinton, who described himself as a “New Democrat”, was a fiscal conservative in disguise, determined to reduce the budget deficit to help lower interest rates and encourage business investment. The US government ran a budget surplus from 1998 until 2001, well beyond the end of his second term, for the first (and last) time since 1969. When the Dot-Com bubble popped in March 2000 (although it took another quarter before market’s realisation set it), the annual “savings” had grown to 1.38% of GDP primarily on the back of spending restraint, effectively withdrawing liquidity from the economy. Conversely, at the end of September, the budget deficit for the first 11 months of the fiscal year 2025 stood at 1.78 trillion USD, 41 billion USD lower than the previous year due to a record 195 billion USD in “net customs receipts”, that is, tariffs. While it was the first time the annual deficit had fallen since 2022, it was still equal to 5.82% of GDP, as outlays climbed to a new all-time high. Finally, back then monetary policy was moving in the opposite direction. Between June 1999 and May 2000 (two months after the market top), the FED hiked its target rate by 175 basis points, from 4.75% to 6.5%. Remarkably, its last increase was a jumbo 50 basis points increase delivered with the Nasdaq Composite Index down almost 30% off its peak.
We posit that what most clearly distinguishes today’s environment from that of a quarter century ago is the pervasiveness of the so-called “wealth effect”. Between March and November 2021, the US entered one of the mildest and shortest recessions in history, shedding about 1% of GDP at most. In our opinion, the economy could have withstood an almost -40% rout in the value of the S&P 500 without entering a severe downturn, as financial assets were playing a far smaller role in household balance sheets. Today, equity valuations relative to GDP are once again at record highs. According to data calculated by the FED, rising equity and property prices have lifted US household net worth to approximately 176.3 trillion USD at the of June 2025, about 5.8 times GDP, with the wealthiest 10% of households responsible for about half of total consumption. As long as asset prices rise, consumer spending remains resilient, even in the face of inflation, tariff shocks and restrictive immigration policy. This dynamic has made the stock market “too big to fail” in a literal sense. A deep equity correction would now translate quickly into a sharp contraction in demand and employment, forcing authorities to intervene. The wealth effect thus acts both as a stabiliser and a risk amplifier, keeping growth resilient while making the economy more sensitive to market drawdowns. Policymakers are aware of this interdependence and have effectively tied financial stability to asset price stability. Government deficits, corporate capex on AI infrastructure and a dovish central bank collectively reinforce this upward loop. Liquidity injections support valuations, those valuations feed household confidence and consumption and that consumption sustains growth. Investors increasingly recognise that in this environment owning assets is not optional, but a prerequisite for keeping up, as their ballooning value becomes a key determinant of prices (and inflation) in the real economy, supplanting labour income. As a result, they chase further investments into stocks, real estate, precious metals and even crypto currencies, perpetuating the very cycle that policymakers depend on. Wall Street has recently picked up on the trend by coining the new buzzword “debasement trade”, which describes the hoarding of real assets to participate in the rise of nominal wealth while real purchasing power falls behind over time. Those factors create a systemic support that didn’t exist during the Dot-Com era. All public and private economic agents are incentivised to sustain asset prices, particularly in strategic sectors like AI and semiconductors. This doesn’t eliminate risk or overvaluation, but it changes the mechanics of how and when corrections occur. The AI boom may well be a craze, but the conditions that would normally end such a cycle are absent.
In this context, it’s difficult to argue for an imminent breakdown or a violent rotation back to value. More likely, we’ll see periodic bouts of volatility rather than a systemic unwind. For investors, the challenge is less about predicting when AI will “burst” and more about managing timing, concentration risk and macro exposure. For this reason, we remain constructive on equities over the medium term. We also believe that international diversification is increasingly important. After having adjusted our positioning in emerging markets, we are shifting our attention toward small caps as an attractive option to gather diversification away from the large-cap momentum trade. In addition, that segment of the market typically performs well in environments where central banks are cutting interest rates and when the economic backdrop is pro-cyclical and supportive of growth. In conclusion, the thesis for being cautious on earnings quality is sound, but the macro backdrop argues against fighting the trend too soon.
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. 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.