AfterValue

So if OpenAi is able to lower its token costs while mainly hosted in Azure, does that not mean, whatever $MSFT deploys becoming cost efficient? The worry was Microsoft doesnt have its competitive inhouse chip like Google TPU or Amazon Graviton. then, how is OpenAi able to provide better token optimization than Anthropics fable which is mainly in AWS?

@negligible_cap thats cheap. BoD will reject rhe offer

*STRIPE, ADVENT OFFER TO BUY PAYPAL FOR $60.50/SHARE: REUTERS $PYPL woah

Bloomberg intelligence estimates $MSFT 1. AI arr will top 90bln in 2027 vs 37bln recently announced 2. Azure will surpass AWS 🤯

I think in football, referee is archaic. There shouldnt be any referee. Technology does far better job. if we left the decision making to technology, it would be much more fairer games. #worldcup

@StockMarketNerd and $IBM will be fine too. They lasted this long because of their deep relationships. They will use AI to better serve. Clients and their employees always want the risk delegated. $IBM delivers on that promise. short term hiccup.

Refreshing to see software not react so horribly to that $IBM infrastructure deal delay headline. That news would’ve sent the entire sector 5% lower 6 months ago. People finally starting to rightfully see these as different companies. Not painting things with such a one size fits all brush.

$IBM has different tech stack vs $MSFT . Saying all saas is exposed based on todays IBM news is just ridiculous.

@mikeschechter bing needs faster response and BETTER sleek UI

The Bing team is partnering deeply with Windows to drive significant improvements to the Windows Search Box - I’m excited for users to start experiencing both the user facing and the under the hood improvements that folks have been heads down on.

there is no way anyone else can outcompete this $MSFT

$MSFT FCF is expected to nearly triple from ~$55B in 2027 to ~$165B by 2030 as Copilot scales across its existing enterprise customer base. That gives Microsoft a clearer path than most hyperscalers to turn heavy AI infrastructure spending into high-margin AI revenue.

@nikesharora @surispeakss @satyanadella there is no need for frontier models though for many of the tasks. Now token consumption from Chinese models have surpassed US ones. Satya would get kudos if he made MAI models available to tweak the weights. Otherwise, I dont get his take.

@surispeakss @satyanadella Maybe because they don't own a frontier model?

I love @satyanadella his ability to put technology transformations in simple terms is par none. In this instance I think it's important to understand the distinction in consumer, horizontal multi-tenant AIaaS and Enterprise which tends to be deep and vertical. 1. Consumer - this is not new news, the consumer has always been part of the product. Be it search (Bing or Google), social media (think TikTok, Snap, and of course FB), even unassuming products like Maps have always retained the knowledge learnt from customers, customers suspecting or unsuspecting have always had the choice. Use the product and share your prompts, location or preferences and that will be used to build a better product. So why is that surprising if it happens in AI, the model complex will continue to use consumer usage to train fundamental multiple modalities. This is the biggest technological event of our lifetime :). 2. Horizontal AIaaS (AI capability that doesn't need to be too enterprise specific - can be tuned, but is a 80% common use case) - Think coding, legal, many current SaaS categories - most likely agentic development for prosumers. All of this behavior is being used to train the model complex to get better at all these. The large swath of small medium size businesses will be fertile training grounds for such applications. They cannot deal with isolated apps and custom deployments. 3. Enterprise deployments - this is where I am not sure the reverse information paradox applies. There is an existing model of isolated single tenant public cloud deployments, deployments where our data, code and connectivity are both isolated and in the hands of the enterprise. This is the deployment we have for our development from all frontier models. All our grounding data, prompts, internal tribal knowledge is sequestered. This is important because this is where enterprise IP will reside. It will be a large task to capture, collate, interpret this data. Equally complex to maintain, evolve and make effective an enterprise AI architecture. But that is what we all will need to sign up for. This is not a cloud vs on prem debate, they can both be equally secured. On prem is probably more unwieldy at the moment given the fast pace of development.

@nicbstme will MAI models address this?

Large neural networks are a bit like black holes for know-how. The more useful you want them to be, the more you feed them: your workflows, your corrections, your evals, your judgment, your taste, your edge… And once that knowledge is absorbed into someone else’s learning loop, the model can start replicating the intelligence without needing the people or organization that created it in the first place. That is the core issue Satya is addressing. In the AI era, companies don’t just accumulate data. They accumulate learning. The real IP is no longer only in documents, databases, patents, or codebases. It is in the traces: prompts, feedback, corrections, evals, agent actions, memories, adapted weights. Every correction is a tiny transfer of human capital into token capital. This is why enterprises will care so much about owning their learning loop. Not just “is my data protected?” but: who owns the intelligence created when my people use AI every day? Because today in consuming intelligence, you are creating intelligence. And what you create should belong to you. Not your weights, not your edge.

@BradSmi thats why Microsoft must lean in heavily towards its own models and make it available for customers to tweak as they wish

Every new generation of digital technology creates a new generation of IP issues. The "Reverse Information Paradox," as Satya describes here, may well create even broader and more profound issues than we've seen in recent decades. It will be vital for businesses and jobs across the economy that we discuss and address these effectively.

@satyanadella AI models should not be black box. If Microsoft deploys MAI model, it should allow the customers to tweak the weights as they see fit for their own needs

@michaeljburry “we love losing money “ one hyperscaler said to another. All rose up and concurred. They decided to lose more

@alexandr_wang if you are benchmarking your LLM based on stock price, I think that is terrible news for $META

new benchmark just dropped 🎁

@DeItaone 0.7pp is margin of error. Better stick to 60/40 for many

AI TAKES THE WHEEL ON WALL STREET JPMorgan is testing AI agents that independently shift investments between stocks and bonds as market conditions change. In 20-year backtests, its top model beat the traditional 60/40 portfolio by 0.7 percentage points annually, with lower volatility. All eight AI agents delivered stronger risk-adjusted returns. But JPMorgan urges caution: the results are simulations, not live performance. AI could also increase crowded trades and amplify market stress.

SpaceX needs to accomplish all these simultanously each year in order to be priced at 205$ PT. Add margin of safety 30% for valuation - you get 143$. Now, assume what is the probability of SpaceX reaching these milestones without any hiccups (regulatory approvals, engineering challenges, demand, customer retention etc), say 50% probability. You have sub 100$ PT comp $SPCX

Basically this is what is projected $SPCX this is a scifi story to me. (download it to be able to zoom in)