“99% of AI Startups Will Be Dead by 2026 — Here’s Why” – a Medium article declared, connecting the exuberance of modern AI with the dotcom era which ended up badly.

My Response: I am not sure this thesis is new: every technology evolution will have winners and category-kings. In 1907, US Steel was the most valuable company in America as steel was the core tech. In 1957, IBM was the king as the mainframe ran the show. By 1983, GE ruled and this era has so far been ruled by Apple and Microsoft on valuations.

But in the midst of those kings, there were many failed companies. At a time, there were dozens of car companies in Detroit with Ford, GM, and Chrysler. How many EV cars failed in the age of Tesla. And as Fairchild Semiconductor evolved with Shockley, semiconductor companies mushroomed but only winners like Intel survived. As I was researching my book – Nanotechnology and Microelectronics – which received the IGI Global 2010 Book of the Year award, I noticed one thing: men lost companies but all created Silicon Valley!

Yes, in the midst of the miry clay, technology empires rose and investors participated, and ego, owo, kudi and money were made. Nigeria did not participate and it is not better than America which did, and lost money in some and won in some. So, if someone will run away from AI because companies will fail, what is new about that? We have that in Africa where because of the fear of risk, we cannot even build boreholes in villages to avoid offending the gods of soil!

In the Igbo Nation, it takes the killing of one leopard to be called a killer of leopards. The deal is the power law of venture investing – the power law describes the principle that a small percentage of investments generate the majority of returns; this means that a few exceptional deals can drive the overall portfolio performance, often outweighing the returns from other, less successful investments –  and that is what most global investors are going after in AI.

AI is not the same as dotcom because AI companies are generating revenue. As YCombinator recently noted, these AI companies are the fastest growing companies on record, and the margins are the best ever. So, even if you agree that 99% will die, the fact is that you are seeing startups which have raised say $300k unlike in the dotcom era where $millions were required to buy HP 9000 server series before the cloud era. So, the collapse of most AI companies would be marginal because many are cheaply funded. Yes, the loss of 1,000 AI companies (not foundation model makers) will be less than just one of those dotcom era companies on financial impacts!

If you are investing in AI companies, look for those which cannot be automated out easily, by making sure they are solving meatspace-level frictions in the physical world. Or they are full-stack AI firms within industrial categories; AI company that offers insurance, banking, legal services, etc and not just selling tech to those sectors

At the Build 2025 developer conference in Seattle, Microsoft laid out an ambitious roadmap for the future of artificial intelligence, one that imagines a world where AI agents from different companies can not only work together but also retain meaningful memories of their interactions with users.

The initiative is a clear signal of Microsoft’s determination to dominate the next phase of the AI revolution — one defined by interoperable, memory-rich digital assistants.

Speaking to reporters and analysts at the company’s Redmond headquarters ahead of the event, Microsoft Chief Technology Officer Kevin Scott said the tech giant is championing open standards that would allow AI agents from different developers to collaborate across platforms. He pointed to the Model Context Protocol (MCP), an open-source standard introduced by Anthropic and backed by Google, as a key part of this strategy.

“Your imagination gets to drive what the agentic web becomes,” Scott said, “not just a handful of companies that happen to see some of these problems first.”

From the Internet to the Agentic Web

Scott drew parallels between today’s AI evolution and the early days of the internet. Just as hypertext protocols in the 1990s unlocked the World Wide Web, Microsoft sees MCP as the infrastructure for what it calls the “agentic web” — a new digital ecosystem where intelligent agents interact, delegate tasks, and co-exist in real time across corporate boundaries.

In this new framework, AI agents would be able to coordinate on tasks like resolving software bugs or managing workflow without being siloed by company firewalls or closed APIs. Microsoft believes this open architecture will democratize AI development and unleash a wave of innovation similar to what the internet enabled three decades ago.

Giving AI a Better Memory

Beyond interoperability, Microsoft is also aiming to solve one of AI’s most persistent limitations: memory. Most of today’s AI agents operate in a transactional, stateless manner — each request or interaction is treated in isolation, often without recollection of previous context. Scott acknowledged this flaw, saying, “Most of what we’re building feels very transactional.”

To address this, Microsoft is introducing a method called structured retrieval augmentation. Instead of attempting to store and recall entire conversations, a process that consumes vast amounts of computing power, this technique allows AI agents to extract and preserve short, meaningful snippets from each interaction. The idea is to create a compact, retrievable roadmap of what was discussed.

“This is a core part of how you train a biological brain,” Scott said. “You don’t brute force everything in your head every time you need to solve a particular problem.”

The approach promises to make AI agents more personalized and context-aware, without the prohibitively high cost associated with current large-context models.

Tools for the AI Workforce

The company is also expected to debut new developer tools to manage and deploy AI agents as part of its enterprise offerings. According to an internal memo reported by Business Insider, Microsoft will roll out “Tenant Copilot” — a system designed to oversee agent activities on behalf of organizations — and “Agent Factory,” a development environment that lets companies build and train their own custom AI agents.

These efforts are part of Microsoft’s broader strategy to position AI agents as “digital teammates” — assistants that go beyond simple chatbot functions to perform sophisticated, autonomous tasks alongside human workers.

By aligning with MCP and pursuing cross-platform standards, Microsoft is signaling that it wants AI to be open, modular, and deeply integrated into everyday workflows, not a set of isolated tools owned by a few tech giants.

The Build 2025 conference, which began on May 19, features keynotes from CEO Satya Nadella and other executives, focusing on Microsoft’s expanding suite of AI technologies. The event is available online, with the company expected to reveal more details about Windows Copilot, AI-integrated development environments, and new partnerships in the AI ecosystem.

However, Microsoft’s big bet rests on whether the industry will embrace its call for collaboration. If successful, the agentic web could reshape how we use and interact with artificial intelligence in the future.

U.S. Treasury yields surged on Monday after Moody’s Investors Service downgraded the country’s credit rating for the first time in decades, sending tremors through the bond market and reigniting debate about America’s worsening fiscal trajectory.

The move came just as House Republicans advanced a Trump-backed tax and spending bill that could add trillions more to the national deficit.

The yield on the 30-year Treasury note jumped 13 basis points to 5.03%, breaking through a psychological threshold that has rattled investors in recent weeks. The 10-year yield climbed 11 basis points to 4.552%, while the 2-year yield rose by 4 basis points to 4.021%. Yields move inversely to prices, meaning the spike reflected a sharp sell-off in U.S. government bonds.

The reaction came swiftly after Moody’s downgraded the U.S. sovereign rating from Aaa to Aa1 on Friday — the first time it has lowered the country’s rating since it began assigning one in 1949. It cited the government’s chronic inability to rein in its budget deficits and the mounting costs of servicing its debt in a high-interest rate environment.

“This one-notch downgrade on our 21-notch rating scale reflects the increase over more than a decade in government debt and interest payment ratios to levels that are significantly higher than similarly rated sovereigns,” Moody’s said in a statement.

The agency said the downgrade also reflects growing doubts over whether any meaningful fiscal reform will emerge from Washington in the foreseeable future.

Moody’s had been the last of the three major ratings firms to keep the U.S. at its top rating, after S&P stripped the country of its AAA status in 2011 and Fitch followed suit earlier this year. Now, all three have relegated the United States to second-tier credit standing — a symbolic blow to America’s long-standing position as the world’s most trusted borrower.

Deutsche Bank analysts called the downgrade “a major symbolic move,” noting that Moody’s had held the line for over seven decades.

But symbolism aside, the timing of the downgrade has concrete implications. It came just as House Republicans pushed through the Trump-endorsed tax and spending package in the House Budget Committee. The bill, which aims to extend tax cuts from 2017 and introduce new tariff policies, is estimated to deepen the deficit by trillions over the next decade.

Moody’s took direct aim at that plan in its downgrade report, warning that “successive U.S. administrations and Congress have failed to agree on measures to reverse the trend of large annual fiscal deficits and growing interest costs.” The agency expressed little faith that current legislative proposals, including those advanced by House Republicans, would meaningfully reduce spending or stabilize debt.

Bank of America economist Aditya Bhave echoed those concerns, writing in a note, “With tax cuts and tariffs hanging in the balance, Moody’s appears to be sending a message that it thinks these policy changes will, on net, put the US on an even worse fiscal trajectory. That is, tariff revenues won’t fully offset the cost of the proposed tax bill. We agree.”

The downgrade and the spike in yields have also shaken confidence in U.S. Treasurys as a safe haven for global investors. For decades, U.S. government debt has been considered the gold standard in global finance — a refuge in times of uncertainty. But with interest payments ballooning and political gridlock worsening, investors are beginning to ask whether that assumption still holds.

Even as debt issuance climbs, foreign demand is showing signs of strain. The Federal Reserve, which had long played a stabilizing role by purchasing Treasurys in the secondary market, has pulled back amid efforts to tighten monetary policy and curb inflation.

Eyes are now on the Fed. Central bank officials — including Atlanta Fed President Raphael Bostic, New York Fed President John Williams, and Dallas Fed President Lorie Logan — are expected to speak on Monday. Markets are watching closely for any shift in tone or hints about how the central bank views the risks stemming from fiscal policy and bond market volatility.

The downgrade also throws fresh scrutiny on the political calculus behind the Republican bill. Trump allies argue that making the 2017 tax cuts permanent is essential to sustaining economic growth and reindustrializing America through tariffs on foreign competitors. But critics say the plan is fiscally reckless, ignoring the weight of soaring debt and the implications of rising borrowing costs.

The national debt now exceeds $34 trillion, and annual interest payments on that debt are approaching $1 trillion, surpassing what the country spends on defense or Medicare. With no bipartisan agreement in sight, analysts warn that America’s fiscal credibility may face even greater tests in the months ahead.