{"id":9553,"date":"2026-10-05T10:50:06","date_gmt":"2026-10-05T09:50:06","guid":{"rendered":"https:\/\/eriya.ch\/?p=9553"},"modified":"2026-10-06T10:20:43","modified_gmt":"2026-10-06T09:20:43","slug":"strategy-update-october-2026","status":"publish","type":"post","link":"https:\/\/eriya.ch\/en\/strategy-update-october-2026\/","title":{"rendered":"Strategy Update October 2026"},"content":{"rendered":"

[vc_row css=”.vc_custom_1754423740115{margin-bottom: 50px !important;}”][vc_column offset=”vc_col-lg-offset-1 vc_col-lg-10″][vc_column_text css=””]<\/p>\n

Strategy Update: October 2026<\/h3>\n

<\/p>\n

The 1000 Billion Dollar Bet<\/span><\/h1>\n

[\/vc_column_text][\/vc_column][\/vc_row][vc_row css=”.vc_custom_1754423773672{margin-bottom: 50px !important;}”][vc_column][vc_btn title=” Prefer to read it as a PDF? Download here.” style=”flat” shape=”square” color=”default” size=”lg” align=”center” i_align=”right” i_icon_fontawesome=”fa fa-regular fa-file-pdf” css=”” add_icon=”true” link=”url:https%3A%2F%2Feriya.ch%2Fwp-content%2Fuploads%2F2026%2F10%2FStrategie-Update-10-2026.pdf|target:_blank|”][\/vc_column][\/vc_row][vc_row][vc_column][vc_column_text css=””]<\/p>\n

REVIEW<\/h3>\n

<\/p>\n

THE FINANCIAL MARKETS IN SEPTEMBER<\/span><\/h4>\n

[\/vc_column_text][\/vc_column][\/vc_row][vc_row css=”.vc_custom_1752586437367{margin-bottom: 80px !important;}”][vc_column offset=”vc_col-lg-offset-1 vc_col-lg-10″][vc_column_text css=””]<\/p>\n

September answered a question that has been preoccupying markets for months: is inflation temporary or persistent? Tokyo set the tone. On the first trading day of the month, the yield on ten-year Japanese government bonds rose to 3%, for the first time since 1996. With that, the last major anchor of low interest rates has come loose. For decades, Japan formed the floor of the global interest rate landscape (alongside Switzerland) \u2013 that floor no longer exists. The consequences were felt around the globe: bond yields rose from Tokyo via London and Frankfurt all the way to Washington.<\/p>\n

<\/p>\n

On 10 September, the European Central Bank raised its key interest rate by a quarter of a percentage point \u2013 already the second hike since the outbreak of the conflict in the Middle East. Six days later, the Fed followed suit: for the first time since summer 2023, it raised its key rate, unanimously, to 3.75% to 4%. Kevin Warsh justified the move with a strong economy, inflation that refuses to come down and rising geopolitical tensions. He left open whether further steps will follow \u2013 but he did not rule them out. Only the Swiss National Bank stayed at zero percent. With inflation well below one percent, it can afford to, and the recently weaker Swiss franc gives it additional help.<\/p>\n

<\/p>\n

The bond market, however, was well ahead of the central banks. The ten-year US yield climbed above 5.2%, its highest level since 2007, and the thirty-year yield at times rose above 5.5%. American households need to prepare for tougher times: 30-year mortgages, the most common type in the US, once again cost more than 7% \u2013 the highest level in almost two years. Anyone buying a house today pays noticeably more interest on the same loan than two years ago.<\/p>\n

<\/p>\n

The driver of the persistent price pressure remains energy. The Strait of Hormuz is still not reliably open, and talks between Washington and Tehran have stalled. Brent crude rose further and is above USD 100, more than 60% higher than at the start of the year. The US government once again released oil from its strategic reserves, which dampens the price but does not solve the problem.<\/p>\n

<\/p>\n

Equity markets proved remarkably resilient. The S&P 500 is only just below its record high and up more than 12% since the start of the year. Beneath the surface, however, the picture was far more divided: only technology, communication services and healthcare gained, while eight of eleven sectors lost ground. The market is therefore not broadening but narrowing \u2013 and depends even more on a small number of technology stocks.<\/p>\n

<\/p>\n

Just how well technology companies are doing is shown by Nvidia\u2019s share buyback programme. At the end of September, the board approved an increase of USD 150 billion \u2013 the largest buyback programme in US corporate history. It is made possible by a profit machine without equal: net income is expected to more than double to around USD 230 billion in the current financial year \u2013 more than all SMI companies earn combined. While the hyperscalers are increasingly taking on debt to fund the build-out, the most important shovel seller is handing its cash back to shareholders. Large buybacks, however, are often also a hallmark of mature companies with slowing growth, and the share has indeed lost momentum in 2026.<\/p>\n

[\/vc_column_text][\/vc_column][\/vc_row][vc_row][vc_column css_animation=”fadeIn” offset=”vc_col-lg-offset-1 vc_col-lg-10″][vc_column_text css=””]<\/p>\n

OUTLOOK<\/h3>\n

<\/p>\n

THE TRILLION-DOLLAR BET IS GETTING MORE EXPENSIVE<\/span><\/h4>\n

[\/vc_column_text][\/vc_column][\/vc_row][vc_row css=”.vc_custom_1752586443552{margin-bottom: 80px !important;}”][vc_column css_animation=”fadeIn” offset=”vc_col-lg-offset-1 vc_col-lg-10″][vc_column_text css=””]<\/p>\n

For months, we have been describing a market here that trades two prices for the same future. The equity side is celebrating record profits, while the long end of the yield curve is pricing in scarcity, deficits and persistent inflation. In September, this divergence did not resolve itself \u2013 it took on a new quality: the central banks have sided with the bond market.<\/p>\n

<\/p>\n

For investors, this marks a turning point. For years, the pattern held: when things get serious, the central bank cuts rates. That safety net no longer exists for now. As long as Hormuz is not reliably open, energy will stay expensive, and as long as energy stays expensive, inflation will remain elevated. The futures curve continues to price in a rapid normalisation of the oil price. Historically, however, quick and clean declines after a genuine supply shock are the exception, not the rule. We are once again living in a world where capital has a cost.<\/p>\n

<\/p>\n

This hits the very core of the current upswing: investment in artificial intelligence. As we show in our focus topic, the large technology companies are spending USD 700 to 800 billion on data centres this year, and more than one trillion for the first time in 2027. Until now, this has largely been paid for out of their own profits. That is changing: around a third of the spending now comes from borrowed money, plus lease commitments of around USD 1.2 trillion off the balance sheet. Every percentage point of higher interest rates makes this bet more expensive. At the same time, revenues are lagging far behind spending, and the big wave of depreciation on the chips already purchased is still to come. The very element whose absence Howard Marks cited for years against the bubble thesis \u2013 high levels of debt \u2013 is now in place.<\/p>\n

<\/p>\n

Growth stocks are particularly sensitive to interest rates because a large part of their value rests on profits in the distant future. When rates rise, those future profits are worth less today. The fact that technology stocks nevertheless gained in September shows how much confidence is already priced in. And the ten largest companies account for almost 40% of the US stock market. Anyone who owns a broad index fund is therefore largely holding a single bet \u2013 increasingly financed with borrowed money, at rising interest rates.<\/p>\n

<\/p>\n

To be fair, the numbers have so far proved the optimists right: the profits are real, and the largest companies have higher margins and stronger balance sheets than 25 years ago. That justifies a higher valuation. However, the earnings estimates for 2027 are missing three costs that have long been real: expensive energy, expensive money and rising depreciation. Sooner or later, all of this will be reflected \u2013 usually first in prices, then in estimates.<\/p>\n

<\/p>\n

When it comes to AI investments, it pays to look at the balance sheet. Companies that fund the build-out from their own stable profits are far better positioned in an environment of rising rates than those increasingly financing it with bonds. Anyone investing via broad indices should know how large their AI concentration really is, and take deliberate countermeasures. We continue to consider real assets and inflation protection attractive. And we treat hedging as what it is \u2013 cheap insurance in a market that prices risks as if they had already been resolved.<\/p>\n

[\/vc_column_text][\/vc_column][\/vc_row][vc_row][vc_column css_animation=”fadeIn” offset=”vc_col-lg-offset-1 vc_col-lg-10″ css=”.vc_custom_1754427734630{margin-bottom: 20px !important;}”][vc_column_text css=””]<\/p>\n

FOCUS<\/h3>\n

<\/p>\n

THE 1OOO BILLION DOLLAR BET<\/span><\/h4>\n

[\/vc_column_text][\/vc_column][\/vc_row][vc_row][vc_column offset=”vc_col-lg-offset-1 vc_col-lg-10″][vc_column_text css_animation=”fadeIn” css=””]<\/p>\n

Never before has so much money flowed so quickly into a single technology. Data centres are springing up everywhere, chipmakers are reporting records, and the US economy increasingly depends on a single investment cycle. But behind the impressive numbers lie uncomfortable questions: who is actually making money from it? Does it all pay off? And what happens when the excavators leave? We show why the AI boom is real \u2013 and so are the risks.<\/p>\n

 <\/p>\n

THE SCALE<\/p>\n

<\/p>\n

The large technology companies, the so-called hyperscalers \u2013 Amazon, Microsoft, Alphabet, Meta and Oracle \u2013 together invested around USD 150 billion in their infrastructure in 2022. For 2026, the market expects USD 700 to 800 billion, and for 2027, more than one trillion US dollars for the first time. In a single year.<\/p>\n

<\/p>\n

A trillion is a number at which the imagination gives up. Spending USD 1 million every day, it would take around 2,700 years to reach one trillion. You would have had to start in the days of ancient Greece \u2013 and Socrates would never have seen the result.<\/p>\n

[\/vc_column_text][\/vc_column][\/vc_row]
\n[vc_row][vc_column offset=”vc_col-lg-offset-1 vc_col-lg-10″][vc_column_text css_animation=”fadeIn” css=””]<\/p>\n

THE US GROWTH ENGINE<\/p>\n

<\/p>\n

Investment in data centres, servers and networks now accounts for around 1.4% of US economic output \u2013 twice as much as a few years ago. Because the rest of the economy is growing only moderately, it delivers, depending on the calculation, around one third to one half of total US growth. Put bluntly: without the AI boom, the world\u2019s largest economy would be growing only half as fast today.<\/p>\n

<\/p>\n

This is visible on construction sites. Spending on data centre construction exceeded USD 50 billion a year for the first time in spring 2026, an increase of 28%. More money is now being spent on buildings for servers than on offices for people. Electricians and HVAC technicians are more in demand than ever, and wages in the skilled trades have risen by around 30% since 2022.<\/p>\n

 <\/p>\n

THE WINNERS<\/p>\n

<\/p>\n

This boom is being paid for by a handful of companies that can simply afford it. Over the past twelve months, the five hyperscalers earned around USD 600 billion combined \u2013 almost five times as much as all the companies in the Swiss Market Index put together. To be fair, part of this comes from valuation gains on stakes in young AI companies. In other words, they also profit from their own customers becoming more valuable. An elegant cycle \u2013 as long as it keeps turning.<\/p>\n

<\/p>\n

The build-out is funded from core businesses. At Amazon, the cloud division AWS accounts for only 18% of revenue but delivers 57% of operating profit. Google pays for its data centres with its advertising business, Microsoft with software and cloud. Put simply: every search query, every Office subscription and every parcel ordered online helps finance a little piece of a data centre. The hyperscalers therefore stand on several legs. How long those legs can carry the growing load is another question.<\/p>\n

<\/p>\n

A second beneficiary is often overlooked: energy. Data centres consumed around 4% of US electricity in 2023; by 2028, this is likely to be 7 to 12%. Worldwide, their consumption will double by 2030 to as much as the whole of Japan uses today. This creates opportunities for utilities and grid operators, but also marks the first hard limit of the boom: chips can be ordered in months, power plants take years.<\/p>\n

 <\/p>\n

WHO REALLY MAKES MONEY?<\/p>\n

<\/p>\n

The big question is not who is investing, but who is making money from it. The answer is clear: the chipmakers. Around 80% of the total gross profit in the AI ecosystem ends up with semiconductor companies. In the traditional cloud software business, the figure is a mere 6%. In August, Nvidia reported quarterly revenue of USD 96 billion \u2013 up 106%, with a gross margin of 75%. Numbers that would make even luxury goods makers envious.<\/p>\n

<\/p>\n

It is the old story of the gold rush: it was not the gold diggers who got rich, but the merchants who sold them shovels and jeans. Levi Strauss sends his regards. Today\u2019s gold diggers are called OpenAI, Anthropic and co. OpenAI expects cash outflows of around USD 280 billion by 2030, and Elon Musk\u2019s AI division posted a loss of more than USD 6 billion in 2025.<\/p>\n

<\/p>\n

The roles, however, are not quite so clear-cut. The hyperscalers do not want to pay Nvidia\u2019s margins forever and are developing their own chips. Nvidia, in turn, is building upwards from the chip \u2013 most recently with the announced acquisition of Hugging Face, a kind of \u201cGitHub for artificial intelligence\u201d. Nvidia\u2019s most important customers are thus also its most dangerous future competitors. You buy from the shovel seller \u2013 while already building your own shovel factory in the backyard.<\/p>\n

[\/vc_column_text][\/vc_column][\/vc_row]
\n[vc_row][vc_column offset=”vc_col-lg-offset-1 vc_col-lg-10″][vc_column_text css_animation=”fadeIn” css=””]<\/p>\n

THE BILL<\/p>\n

<\/p>\n

Does it all pay off? The honest answer: not yet. AI revenues now cover the ongoing depreciation on data centre hardware, but only around half of the cumulative bill. The consultancy Bain estimates that around USD 2 trillion in annual AI revenue would be needed by 2030 to justify the build-out. Today, depending on the definition, the figure is USD 50 to 150 billion. The gap is not small \u2013 it is enormous.<\/p>\n

<\/p>\n

On top of that, there is a ticking time bomb in the accounts. In the twelve months to March 2026, the four largest hyperscalers invested around USD 434 billion but booked only around USD 149 billion in depreciation. Today\u2019s profits therefore reflect only a fraction of the actual costs. The wave of depreciation is coming \u2013 whether revenues keep pace or not.<\/p>\n

 <\/p>\n

ON BORROWED MONEY<\/p>\n

<\/p>\n

As long as profits keep flowing, this is not a problem. But investment is growing around three times as fast as operating cash flow. Around the third quarter of 2026, the two lines cross \u2013 from then on, self-generated cash is no longer enough. The build-out is increasingly being financed with debt.<\/p>\n

<\/p>\n

Between 2020 and 2024, the five hyperscalers raised an average of around USD 35 billion a year in new bonds. In 2025, it was USD 93 billion, and in 2026 already USD 132 billion by the end of July. Alphabet even issued a bond with a maturity of 100 years. Whoever gets their money back in 2126 will probably no longer remember ChatGPT. On top of that come lease commitments for data centres of around USD 1.2 trillion that do not even appear on the balance sheet. Howard Marks long argued that a real bubble lacked debt. That element is now in place.<\/p>\n

 <\/p>\n

CONCENTRATION RISK<\/p>\n

<\/p>\n

The ten largest companies in the S&P 500 now account for almost 40% of the index \u2013 more than ever in modern stock market history. Anyone who owns a broad US index fund believes they are diversified, but in reality is largely betting on a single wager. Particularly exposed is, of all things, the biggest winner: Nvidia most recently generated USD 89 billion of its USD 96 billion in revenue from its data centre business. As long as investment continues, that is a dream business model. If it slows, hardly anyone will feel it as directly.<\/p>\n

[\/vc_column_text][\/vc_column][\/vc_row]
\n[vc_row][vc_column offset=”vc_col-lg-offset-1 vc_col-lg-10″][vc_column_text css_animation=”fadeIn” css=””]<\/p>\n

WHEN THE EXCAVATORS LEAVE<\/p>\n

<\/p>\n

The jobs boom around data centres is taking place on construction sites. In the US state of Georgia, around 28,350 people are currently working on data centre construction sites, but only 5,471 are permanently employed in the data centres already in operation. A finished data centre is, above all, a building in which machines do the work. They don\u2019t complain \u2013 but they don\u2019t go shopping in the neighbourhood either.<\/p>\n

<\/p>\n

From January to August 2026, US employers attributed more than 116,000 announced job cuts to AI \u2013 around 22% of all announced job cuts. If the build-out phase ends, a double effect looms: the construction workers leave, and offices become leaner. That would hit the American consumer, so far the most reliable pillar of the economy.<\/p>\n

 <\/p>\n

THE DECISIVE QUESTION<\/p>\n

<\/p>\n

In the end, everything hinges on one question: does AI actually make us more productive? Only then will the trillions pay off. The signals are mixed. Labour productivity in the US is rising strongly, whereas total factor productivity, which measures genuine efficiency gains, is barely growing. The Federal Reserve Bank of San Francisco puts the probability of a sustained productivity boom at 57% \u2013 or only 21%, depending on the measure.<\/p>\n

<\/p>\n

A look back gives cause for optimism: in the 1990s, too, only labour productivity showed up at first, and the real efficiency gains followed years later. But back then, the dotcom bubble also came along in between. You can be right about the technology and still lose money on the timing.<\/p>\n

 <\/p>\n

AND EUROPE?<\/p>\n

<\/p>\n

Europe is largely watching from the sidelines. The US has around 75% of global AI computing capacity, the EU less than 5%, and the US productivity lead per hour worked has more than doubled since 2018. That is both a risk and an opportunity: Europe bears hardly any of the costs of the boom, but also benefits less. Yet Europe does not need to build the data centres itself to benefit from the technology \u2013 it just needs to use it.<\/p>\n

[\/vc_column_text][\/vc_column][\/vc_row]
\n[vc_row][vc_column offset=”vc_col-lg-offset-1 vc_col-lg-10″][vc_column_text css_animation=”fadeIn” css=””]<\/p>\n

CONCLUSION<\/p>\n

<\/p>\n

The AI boom is real. The investments are genuine, so are the profits of the largest companies, and the technology will change the economy. But so are the risks. Revenues are lagging far behind spending, financing is shifting from solid balance sheets to debt, and markets depend on fewer and fewer names. For us, AI infrastructure is therefore not a certainty but a bet \u2013 a very big one.<\/p>\n

<\/p>\n

This also affects our portfolios, which include Microsoft, Alphabet and Nvidia. We are therefore watching closely to see who succeeds in conquering new areas of the AI value chain without jeopardising the profitable core businesses that pay for the build-out in the first place. For investors, this means: don\u2019t get out, but don\u2019t put all your eggs in one basket either. Anyone investing via broad indices should know how large their AI concentration really is.<\/p>\n

[\/vc_column_text][\/vc_column][\/vc_row]<\/p>\n<\/div>","protected":false},"excerpt":{"rendered":"

[vc_row css=”.vc_custom_1754423740115{margin-bottom: 50px !important;}”][vc_column offset=”vc_col-lg-offset-1 vc_col-lg-10″][vc_column_text css=””] Strategy Update: October 2026The 1000 Billion Dollar Bet [\/vc_column_text][\/vc_column][\/vc_row][vc_row css=”.vc_custom_1754423773672{margin-bottom: 50px !important;}”][vc_column][vc_btn title=” Prefer to read it as a PDF? Download here.” style=”flat” shape=”square” color=”default” size=”lg” align=”center” i_align=”right” i_icon_fontawesome=”fa fa-regular fa-file-pdf” css=”” add_icon=”true” link=”url:https%3A%2F%2Feriya.ch%2Fwp-content%2Fuploads%2F2026%2F10%2FStrategie-Update-10-2026.pdf|target:_blank|”][\/vc_column][\/vc_row][vc_row][vc_column][vc_column_text css=””] REVIEWTHE FINANCIAL MARKETS IN SEPTEMBER [\/vc_column_text][\/vc_column][\/vc_row][vc_row css=”.vc_custom_1752586437367{margin-bottom: 80px !important;}”][vc_column offset=”vc_col-lg-offset-1 vc_col-lg-10″][vc_column_text css=””] September…<\/p>\n","protected":false},"author":10,"featured_media":9567,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"om_disable_all_campaigns":false,"_monsterinsights_skip_tracking":false,"footnotes":""},"categories":[46],"tags":[],"class_list":["post-9553","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-strategy-updates-en","category-46","description-off"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t