MICROSOFT FY26 and Q1 27 results--Tier 2 holding--successfully threading the needle?
MSFT FY26 and Q1 27—Tier two –successfully threading the needle?
I have to admit to not looking forward to writing this
piece. Why? Because it is not that easy. MSFT is known for strong management and
strong financials, products that are embedded and bundled into must-have
propositions for enterprise customers. That still comes through the numbers and
why, at the right price, it is a core holding. There are a few huge lingering questions
that, IMO, remain unsolved.
The following are those medium-term (?) issues and can go
either way IMO at this stage.
1.
Long-term returns on data centre builds. Perhaps
this is the easiest to understand, but still an unknown. Will DC output (tokens)
be a commodity and earn a cost-of-capital-type return, or will they continue to
generate excess returns? Or will some operators generate excess returns and others
not? ATM they are very attractive investments, and all the hyperscalers are investing
heavily. MSFT has described what they do by offering compute with a software layer
over the top. Any customisation and stickiness could be critical in the LT. Perhaps MSFT is showing the most investment constraint
here; that sort of makes sense. MSFT does not have its own model to feed tokens
into like GOOG, and it also doesn’t have as advanced a chip business as GOOG or
AMZN, so it is most exposed to the NVDA tax. MSFT has been conservative in its
R&D spend, and that has resulted in lagging the others. Will that cost them
in the future? As I look to the layers of vulnerability in this AI structure,
MSFT is certainly not at the bottom, but it is not at the top either.
2.
The evolving OpenAI relationship. MSFT's
relationship and strategy around OpenAI has evolved dramatically, and it is
probably not over. A couple of years ago, MSFT saw itself as the huge beneficiary
of the strongest frontier model, through an exclusive relationship and as an equity
investor. That narrative has now gone. The exclusive and integrated OpenAI
model distribution strategy has been replaced with MSFT proposing itself as a deterministic,
zero trust harness and control panel for customers. MSFT will govern, route, and
audit untrusted models from any vendor. Under this strategy, Foundation models
are hopefully commodities or, nevertheless, not to be trusted by the underlying
customer, with MSFT being the trusted layer between the client and the foundation
models that may steal clients' IP and then replicate and destroy them in time. That
is an understandable but huge somersault in strategy. MSFT also has a couple of
lucrative revenue and profit share agreements with OpenAI, which, importantly, have
drop-dead dates when those earnings streams end. In some regards, MSFT has a
hedge here, as a 27% shareholder in OpenAI; if the foundation models become dominant
in replicating software, and many other things, MSFT gains from that shareholding.
If the foundation models fail in their more exuberant ambitions, MSFT's base business
and position in the stack retains more value. As a shareholder, I don’t like scenario
one as much as two; being a sum-of-the-parts proposition doesn’t sound that attractive
to me.
3.
That logically leads to the last issue, and
perhaps the biggest for MSFT: will its position in the client value chain be
defensible longer term? Through its office set of software products, MSFT earns
enormous margins and participates in reasonable growth. MSFT successfully bundle
strong products with weak products and gains share in the enterprise software market.
MSFT's client relationships are second to none. When looking at the threats of
foundation models, MSFT's response makes perfect sense, and the strategy is
defensible, but is it enough? MSFT's bet, at this stage, is that there will be fragmentation
in foundation models, that being a fast follower will be enough, and that the
large LLMs can't be trusted with client data IP, as they could be a competitor,
while MSFT has proven not to be. That strategy means MSFT moving closer to the clients,
embedding products like Copilot, which may not be leading edge but close
enough, and that they have clients' interests at heart. Ok, that works for now,
but it does not take too much imagination to see the world evolve into a threatening
place. For instance, if there is one preeminent model, or if the frontier
models keep their best models to themselves, allowing them to develop better products
down the stack, in the application layer. The scenarios mushroom out. There is also
healthy competition in the harness and application layer with NOW, CRM, and Anthropic
all eyeing or making moves in this space. The end issue is where the power will
reside in the AI stack, and that is yet to be determined. Execution and positioning
will be important as well, and MSFT is better placed than most. Ironically, I think
a complete washout and cleansing of the system, where access to liquidity dries
up, would be extremely bullish for the hyperscalers and embed their dominant oligopoly
for decades to come. At this stage, it is mainly about watching and seeing how
the power in the AI stack develops and spreading your bets.
FY26 Q1 27 Results and numbers.
The FY was solid to good, as was Q1 27, which was better.
Revenues in FY26 grew 18%, 1% better than my estimates; NPAT was +32%, 7% better, but included OpenAI MTM. Excluding this revaluation of the equity
stake, the NPAT was +22%, about 4% higher than my estimate. The positive came
from operating leverage exceeding expectations. MSFT runs a tight ship and
can operate the businesses very efficiently. Below we see the GM pressure by
quarter and the improving operating margin on an annual basis. The improvement
in GM in Q127 was a welcome relief, with the cost of build having a negative impact,
while MSFT continue to squeeze efficiency, improving operating margins, a positive
story. The MSFT operational mix is a structural negative to op margins, with
Azure and Cloud growing very quickly, while the highest-margin software business
grows more slowly, presenting a structural challenge to margins. The margin
improvement is more impressive given the mix change but will be hard to keep improving.
MSFT is willing to grow lower-margin but high-volume profits.
MSFT has recast their segments, and I am not sure I am a fan
of this; it seems to be applauded by the market. The three previous segments
have been changed to two, with the dominant Agents and Infrastructure being 87%
of income, while the rest is the Consumer and Devices segment, being the legacy
business including gaming. We can probably say Activision Blizzard was not a great
move, and gaming could be shed in time. First-time disclosures of Azure revenues
inside the main division were made. The Q1 27 results showed Agents and Infra growing
revenue 25% and income 24%, part of the structural shift mentioned above. Distinguishing
the difference between this structural shift in mix and the impact of higher depreciation
from huge capex on income could be a challenge going forward, an important
distinction. Azure revenues grew 48% and comprise 43% of that division's revenue
and 36% of the total company. From this we can see a very fast-growing Azure business,
a large, highly profitable but slower-growing software business, and a consumer
stub going nowhere and quickly becoming irrelevant. Q2 guidance basically forecasts
more of the same.
MSFT pointed out the growth in Copilot, which accelerated,
but MSFT did not disclose the most recent figures; it sits at about 8% of the
installed base, and it will be interesting and important to see how well MSFT
can get AI products into the client base. MSFT are changing the pricing
structure to include volume-based pricing, as AI has high variable costs. Important
that this sticks. Foundry (the model reseller business) was mentioned as growing
3X on pcp. The other factor to mention in the results is that profits generated
from OpenAI agreements could be around 10% of FY27 earnings and extend from
2028 to 2032 before terminating, then creating an earnings headwind.
I do not share the current market's negative obsession with
hyperscaler negative FCF due to capex investment. It simply depends on the ROI
on this investment; the financial stability of all the hyperscalers is currently
hardly at risk. In fact, an argument could be made that it is in the hyperscalers'
interest to raise the stakes here and make it a “big boys” game and cement the oligopoly
instead of encouraging a fragmented industry. Currently, ROI is attractive but
should be closely monitored.
The below numbers show a mixed outcome. The high capex
weight on returns, but if we match capex to earnings a year later, the
incremental ROE for MSFT continues to be sound. Note that the backlog is about 45%
dependent on OpenAI, so MSFT still runs an OpenAI concentration risk. If OpenAI
fails, I would expect MSFT to move to control, but that risk remains remote at this
stage, with the funding markets open.
In terms of forward commitments, MSFT sits at (30/9/26) $365B, a large number. With MSFT earning around +$130 B, the obligations are high but nowhere near life-threatening; earning an adequate return remains critical. MSFT added a comment that these commitments start from 2027 to 2034 and are from 1-20 years and are, in some cases, conditional. Any flexibility in contracts would be welcome.
SUMMARY and VALUATION
MSFT is playing to its strengths, as it should do. That strength
is the embedded nature of its position in its enterprise customers' operations.
Currently, that looks sound, but we are in an industry where large and rapid change
can occur. Those changes could be beneficial or not at this stage. The bear
case revolves around its value proposition being supplanted by new competitors.
The two big value propositions here are being the harness layer to control the
agents and models into the enterprise customer base and the returns on the DC
spend. The thesis could weaken if MSFT's position in these areas proves unsustainable.
With these uncertainties and issues, MSFT remains a T2
investment for me. There is little doubt it is one of the best-run companies,
and that should give it some leeway. My base case is that the hyperscalers
prove the winners in this race. They have their own ecosystems, resources and
market positioning, and in most cases the breadth of products. Through this lens,
MSFT sits strong but partly vulnerable.
At $523, and assuming 14% 5Y eps cagr and a 23X exit rate, it
generates a 7% return. A 10% return is generated at $470. The thesis holding is
far more important than the apparent value on offer; they are inherently
connected.






Comments
Post a Comment