Mad Hedge AI Archives

The Algorithmic Abyss: How AI Deregulation Threatens to Plunge Financial Markets into Chaos

Washington D.C. - A seismic shift is underway in the landscape of financial regulation, as burgeoning calls for the deregulation of artificial intelligence (AI) within the sector gain momentum. While proponents tout the potential for innovation and efficiency, a growing chorus of experts warns that unchecked AI deployment could unleash unprecedented volatility and systemic risk upon global financial markets.

The debate, which has intensified following recent policy shifts, centers on the balance between fostering technological advancement and safeguarding the stability of the financial system. Critics argue that the rapid evolution of AI, coupled with insufficient regulatory oversight, creates a fertile ground for market manipulation, algorithmic flash crashes, and a potential erosion of market integrity.

The Deregulation Drive:

The push for AI deregulation is fueled by several factors. Firstly, the financial industry's embrace of AI has accelerated dramatically, with algorithms now playing a crucial role in everything from high-frequency trading to risk assessment and portfolio management. Advocates claim that stringent regulations stifle innovation and hinder the United States' ability to compete in the global AI race.

Secondly, a prevailing sentiment within certain political circles emphasizes minimizing government intervention in the market, viewing regulation as an impediment to economic growth. This ideology, combined with powerful lobbying efforts from tech and financial firms, has created a political climate conducive to deregulation.

"We must unleash the transformative power of AI to drive economic prosperity," argues a prominent industry lobbyist, speaking on condition of anonymity. "Excessive regulation will only serve to hamstring our financial institutions and cede our competitive advantage to nations with more permissive regulatory environments."

However, this perspective is met with fierce opposition from regulators, academics, and consumer advocacy groups, who express deep concerns about the potential consequences of unfettered AI deployment.

The Shadow of Algorithmic Risk:

One of the most pressing concerns revolves around the inherent complexity and opacity of AI algorithms. "These systems are often black boxes," explains Dr. Eleanor Vance, a leading expert in financial risk management. "We don't always fully understand how they arrive at their decisions, which makes it incredibly difficult to anticipate or mitigate potential risks."

This lack of transparency poses a significant challenge for regulators, who are tasked with ensuring market fairness and stability. The potential for algorithmic bias, where AI systems perpetuate or amplify existing inequalities, further complicates the regulatory landscape.

Furthermore, the interconnectedness of AI systems within the financial ecosystem creates the potential for cascading failures. A single algorithmic error or malicious attack could trigger a chain reaction, leading to widespread market disruption and systemic risk.

"We've already seen instances of algorithmic flash crashes, where automated trading systems triggered rapid and dramatic price swings," warns a senior regulatory official. "Without proper safeguards, these events could become far more frequent and severe."

Concerns of Market Manipulation:

The potential for AI-powered market manipulation is another major source of concern. Sophisticated algorithms could be used to exploit market vulnerabilities, engage in predatory trading practices, or spread misinformation to manipulate asset prices.

"Imagine an AI system designed to detect and exploit subtle patterns in market data, allowing it to front-run trades or manipulate prices with unprecedented precision," says a cybersecurity expert specializing in financial systems. "The potential for abuse is immense."

The proliferation of deepfakes and AI-generated misinformation further exacerbates these concerns. Malicious actors could use these technologies to spread false rumors or manipulate market sentiment, creating artificial volatility and profiting from the resulting chaos.

The Regulatory Void:

The current regulatory framework is ill-equipped to address the unique challenges posed by AI. Existing regulations, designed for traditional financial instruments and trading practices, are often inadequate for overseeing complex algorithmic systems.

"We're facing a regulatory gap," admits a financial regulator. "The pace of technological innovation has outstripped our ability to develop effective oversight mechanisms."

The development of new regulatory frameworks is further complicated by the lack of consensus on best practices and ethical guidelines for AI deployment in finance. International cooperation is also crucial, as financial markets are increasingly interconnected, and regulatory arbitrage could lead to a race to the bottom.

The Social and Economic Implications:

The potential consequences of AI-driven market instability extend far beyond the financial sector. A major market crash could trigger a global economic recession, leading to widespread job losses, social unrest, and a loss of public trust in the financial system.

Furthermore, the increasing reliance on AI in financial decision-making raises concerns about algorithmic bias and discrimination. AI systems could perpetuate existing inequalities, denying access to credit or investment opportunities to marginalized communities.

"We need to consider the social and ethical implications of AI deployment in finance," emphasizes a social justice advocate. "These systems should be designed to promote fairness and equity, not to exacerbate existing disparities."

The Call for Responsible Innovation:

Despite the risks, many experts believe that AI has the potential to revolutionize the financial industry, improving efficiency, reducing costs, and expanding access to financial services. However, they stress the need for responsible innovation, guided by robust regulatory oversight and ethical principles.

"We need to strike a balance between fostering innovation and mitigating risk," argues a financial technology expert. "This requires a collaborative effort between regulators, industry leaders, and academic researchers."

Key recommendations include:

Enhanced transparency: Requiring financial institutions to disclose the algorithms used in their trading and risk management systems.
Robust risk management frameworks: Developing new regulatory standards for algorithmic trading and AI-driven financial decision-making.
Ethical guidelines: Establishing clear ethical principles for AI deployment in finance, addressing issues such as bias, discrimination, and accountability.
International cooperation: Harmonizing regulatory standards and best practices across jurisdictions.
Continuous monitoring and evaluation: Establishing mechanisms for ongoing monitoring and evaluation of AI systems to identify and mitigate potential risks.

The Future of Finance:

The future of finance hinges on our ability to navigate the complex challenges posed by AI. A failure to establish robust regulatory safeguards could lead to a period of unprecedented market volatility and systemic risk, with potentially devastating consequences for the global economy.

However, if we can embrace responsible innovation, guided by ethical principles and robust oversight, AI has the potential to transform the financial industry for the better, creating a more efficient, inclusive, and resilient financial system.

The coming years will be critical in determining whether we can harness the power of AI for the benefit of society, or whether we succumb to the algorithmic abyss.

BLOOD-OATH NDAS AND BOURBON-FUELED PREDICTIONS

Mad Hedge AI

(NVDA), (AAPL), (NFLX), (AMD), (IBM), (AMZN), (GOOGL), (META)

I was sipping an overpriced airport cappuccino when Jensen Huang strutted onto the GTC 2025 stage. My seatmate didn't recognize him, but I'd been at a private dinner with one of Nvidia's VPs the night before.

After 40 years covering tech from Tokyo to Wall Street, you build a network that pays dividends in boardroom whispers. "Tomorrow's announcements will make portfolio managers drool," he'd hinted, remembering when I'd steered his family into Apple (AAPL) back in 2003.

Two minutes into Jensen's keynote, my neighbor's jaw dropped faster than Netflix (NFLX) stock after missing subscriber estimates. By our boarding call, he was frantically buying Nvidia (NVDA) shares.

"Too late, kid," I muttered. "You're about $2 trillion late to this party."

While investors obsess over Fed rates, Jensen has built the infrastructure for the biggest gold rush since the internet. Nvidia is selling the only shovels that actually work.

The Blackwell Ultra processor, due late 2025, isn't just an upgrade but a leap in AI reasoning that makes current models look primitive. The Vera Rubin server, arriving in 2026, is expected to be 3.3x faster.

A Stanford classmate who designs power systems texted during the keynote: "Been testing prototypes for months. Had to sign blood-oath NDAs."

When your college buddies design the circulatory systems for the world's digital brain, you get texts analysts would trade their Bloomberg terminals for.

Jensen projected global data center spending reaching $1 trillion by 2030. When I repeated this to my bartender at the St. Regis, he asked if I was having a stroke.

"No one spends a trillion on computers." I reminded him no one used to spend billions streaming cat videos either.

Who benefits from this tsunami?

Nvidia is obvious, with 20-25% projected annual growth looking conservative.

Advanced Micro Devices (AMD) has been gaining ground. At a San Jose conference, I caught an engineer who "accidentally" left next-gen chip data visible.

When you've covered tech since before AMD made math coprocessors for IBM (IBM) clones, you spot when a "mistake" is actually a strategic leak.

Amazon (AMZN) remains the cloud king. During an investor site visit, an AWS director let slip their AI buildout is "significantly ahead of guidance."

My history covering AWS when analysts dismissed it means their PR people don't hover during these visits. They forget I reported on Amazon when Bezos was still shipping books from a garage.

Google (GOOGL) keeps their best tech garaged. In an Uber after a Palo Alto event, a tipsy engineer boasted: "The public sees maybe 20% of what we're running internally."

Twenty years of moderating "Future of Computing" panels puts me in countless rideshares with people building that future, who forget I write for fund managers controlling billions.

Meta (META) is investing heavily in generative AI. At an industry roundtable, their researcher, after his third wine, sketched neural networks on napkins.

Having covered companies from garage stages to trillion-dollar valuations gives you invisibility at these events – they see you as furniture rather than media.

The generative AI market is projected to reach $62.72 billion by 2025, but at a Bloomberg roundtable, fund managers confided they're allocating "multiples" of their usual positions.

After decades of running a hedge fund that spotted Japan's rise, the dot-com boom, and the fracking revolution, you get invited to side conversations where real money moves are discussed.

My advice: don't pick a single winner, think long-term, and watch innovation closely.

My nephew recently asked for stock tips for his summer job earnings. "Buy Nvidia and delete your trading app for 10 years," I advised. He looked offended. "Everyone knows Nvidia. I want something undiscovered."

I patted his shoulder. "Kid, sometimes the biggest mistake in investing is trying to be clever when the obvious answer is staring you in the face."

In the AI revolution, it's not about finding the next Nvidia. It's about not overthinking the Nvidia that's already here.

FORGET THE HERD, FOLLOW THE ENGINEERS

Mad Hedge AI

(MRVL), (MSFT), (NVDA), (META), (AVGO)

Back in 1981, when I was tramping through the backcountry of Japan researching semiconductor factories, a wise old engineer told me something I've never forgotten: "Young man, in technology, the custom always beats the commodity."

Four decades later, watching Marvell Technology's (MRVL) strategic pivot to custom AI silicon, that weathered engineer's words are ringing in my ears.

The recent tech selloff triggered by Microsoft's (MSFT) canceled data center leases sent most chip stocks tumbling. The market, doing what it does best, threw the baby out with the bathwater.

While the herd was stampeding for the exits, I was busy loading up on MRVL shares. Wall Street's short-term memory loss is your gain, folks.

Marvell delivered a knockout 27% year-over-year growth in Q4'25, and that's just the appetizer. The main course is coming as their custom silicon strategy hits full stride.

Let me tell you why this matters.

Remember when Nvidia was just "that gaming chip company" before it became the AI powerhouse worth more than most countries' GDP? Marvell is following a similar trajectory but with a crucial difference - they're not trying to be everything to everyone.

They're the special forces team designing precision instruments for the AI revolution while Broadcom (AVGO) plays the role of the 800-pound gorilla servicing hyperscalers. Different animals, different hunting grounds, both eating very well.

Here's what the nervous nellies are missing: management is betting the farm on data center product development, expecting to blow past their $2.5b AI systems revenue target for eFY26.

When executives redirect R&D dollars like this, they're not guessing – they're following customer purchase orders. I've been in this business long enough to know that's the financial equivalent of smoke signals from the Vatican - big news is coming.

Need more evidence? I've got it straight from the engineering trenches.

During my tech conference rounds last month, three different chip engineers cornered me to rave about Marvell's 800G PAM and 400ZR DCI products.

If that sounds like alphabet soup to you, just know this: they're selling tomorrow's technology while competitors are still perfecting yesterday's.

Their Electro-Optics products and Teralynx Ethernet switches aren't just growing – they're flying off the shelves.

Meanwhile, the Meta Platforms (META) partnership they announced in Q4'25 is pure gold.

With Meta planning to drop $60-65b in eFY25 on Llama 4 development, Marvell just secured a VIP pass to the most expensive tech party of the decade.

So what does all this mean for Marvell's numbers? Let's get concrete.

For Q1'25, I'm forecasting $1.9b in revenue with EPS of $0.62/share. Their margins will expand throughout eFY26 as volumes ramp up.

Financially, Marvell is rock solid. They closed FY24 with $948mm in cash, $4b in debt, and a leverage ratio of 1.58x – their best position since Q4'23. In fact, Fitch upgraded their credit rating to BBB with a stable outlook.

In layman's terms: the books look cleaner than a surgeon's operating room.

Yes, inventories increased to 111 days on hand ($1,030mm). Normally, I'd raise an eyebrow at this. But in today's supply-chain-from-hell environment, extra inventory is like having an extra gas tank in the desert – suddenly the smartest idea ever.

I rang up a hyperscaler CTO friend last week (no names, but his company has more users than most countries have citizens).

His take? "General-purpose computing for AI is like using a sledgehammer to hang a picture. We're moving to custom solutions wherever possible." That's Marvell's sweet spot.

MRVL shares have been beaten like a rented mule, creating an entry point with fair value at $204/share by my analysis. Even my conservative case gets you to $157 – still double today's price.

This reminds me of standing atop Mount Fuji with a Japanese tech CEO in 1978.

Looking down at Tokyo, he explained how his company was shifting from consumer electronics to specialized components the world's biggest brands couldn't manufacture themselves.

Those who spotted that transition early made fortunes. Specialized excellence always commands a premium – in mountaineering, in technology, and certainly in today's custom silicon market.

WHY I'M BREAKING MY OWN RULES FOR THIS AI STOCK

Mad Hedge AI, Uncategorized

(TEM), (ILMN)

Last weekend, my daughter, the computer science whiz studying at UC, called me with a question about healthcare AI companies for her investment project.

"Dad, my professor says these healthcare AI stocks are just cash incinerators with no path to profitability," she explained. "Should I avoid the whole sector?"

I was cutting vegetables for dinner as we talked – my homemade pasta sauce is legendary in three counties. I set down the knife and gave her my perspective, which I'll share with you now.

Back in the early '90s, when I was covering Asian markets for The Economist, I encountered countless biotech companies making grand promises about revolutionizing healthcare.

The pattern was always the same – ambitious visions, heavy cash burn, and perpetually delayed profitability. Few survived.

So when a healthcare AI company crosses my radar these days, my first instinct is skepticism, honed by decades of watching promising technologies evaporate along with investors' capital.

But Tempus AI (TEM) has me breaking my own rules.

You see, there's a moment in every significant technological shift when the numbers start telling a different story than the narrative.

TEM's financials have significantly improved, with positive cash flow expected by year-end and Q4FY24 delivering record revenues.

The Data & Services segment – where the real money is – grew by a stunning 44.6% YoY.

FY24 revenues hit an all-time high of $693.4 million, bolstered by deals with Boehringer Ingelheim and Illumina (ILMN).

Management is now guiding for $1.24 billion in 2025 revenues – a 79% YoY growth. This isn't idle talk – these are numbers that give hardened skeptics like me reason to take notice.

The first question I ask about any AI company is whether they actually have the data to do what they claim.

Tempus AI has amassed over 240 petabytes of healthcare data – an information advantage that reminds me of trading Japanese equities in the 80s when having access to real company data was worth its weight in yen.

In Q4FY24, Tempus released Olivia, their AI-enabled personal health app.

Unlike other health apps that fragment patient information, Olivia consolidates data from multiple providers into a single interface. It gives patients access to their complete health records and delivers AI-powered insights about their conditions.

Having watched numerous healthcare startups flame out during my reporting days, I can tell you this approach solves a genuine problem that most tech companies miss.

For FY24, revenues grew by 30% compared to 77% in FY23.

Don't be fooled by the apparent slowdown – TEM is working from a much larger base, with higher-margin services taking center stage.

The company holds $448 million in cash and short-term investments with a quarterly burn rate of $39 million, giving them a runway of nearly three years.

The Ambry Genetics acquisition is a game-changer for TEM, adding genetic testing capabilities and a cool $300 million in revenue.

Back when I was tramping through biotech labs in Asia for The Economist, I learned a critical lesson - the bottleneck in genetic medicine isn't sequencing but interpretation. TEM knows this and is planting its flag exactly where the gold is.

When Q4 revenues came in a hair below Wall Street's guesses, the stock took a hit. I've seen this movie before.

The algorithms panic, creating beautiful entry points for those of us with enough battle scars to know better.

The CEO barely contained his satisfaction: "Our Data and Services business just had a really strong Q4, finishing a really strong year."

Translation: "We're killing it but I'm not going to brag."

Sure, TEM has rivals. PathAI, Prognos Health, and Healwell AI are all scrambling for a piece of this pie.

But none has TEM's data treasure chest, and none has figured out how to monetize in three directions at once - genomic testing, data licensing, and consumer apps.

The market size is staggering - somewhere between $317 billion and $490 billion by 2032. That's bigger than the GDP of most countries I've reported from.

And here's the kicker with AI companies - the rich get richer. More data attracts more customers, generating even more data. Once you're ahead, staying ahead gets easier.

For those who care about valuation (and you should), I'm looking at 2026 projected revenues of $1,612 million with a P/S multiple of 6.57. That math gives me significant upside from today's price of around $49.87.

Is that multiple too rich? Not when you're dealing with a company that's cracked the code on healthcare AI profitability.

Are there risks? Of course. Profitability might take longer than expected.

The data moat could theoretically be bridged by a determined competitor with deep pockets. The field is getting crowded.

But I’ve witnessed enough market transformations to recognize when a company sits at the perfect intersection of powerful trends. And right now, TEM is riding three unstoppable waves—AI, precision medicine, and healthcare’s digital overhaul.

The recent market jitters have created a textbook buying opportunity. When Wall Street's short-term anxiety gives you a chance to buy long-term winners at a discount, you take it.

TEM has multiple ways to win, and that's the kind of bet I've made my career on.

With that, I had to end our call before my sauce burned. Yesterday, she texted me that she'd bought TEM on the pullback.

After decades navigating markets from Tokyo to Wall Street, there's nothing quite like seeing the next generation apply those lessons – sometimes even better than their teacher.

The student becomes the master, as they say.

Oracle Financial Services Supercharges Investigation Hub with AI Agents, Revolutionizing Financial Crime Detection

Mad Hedge AI

Oracle Financial Services Supercharges Investigation Hub with AI Agents, Revolutionizing Financial Crime Detection

In a significant leap forward for financial crime prevention, Oracle Financial Services has unveiled a major enhancement to its Investigation Hub Cloud Service, integrating a powerful suite of artificial intelligence (AI) agents and agentic workflows. This advancement promises to dramatically accelerate and improve the efficiency of financial institutions' investigative processes, enabling them to uncover complex patterns and combat increasingly sophisticated financial crimes.

The financial sector faces a relentless barrage of threats, from money laundering and fraud to terrorist financing. Traditional investigative methods, often reliant on manual data gathering and analysis, struggle to keep pace with the sheer volume and intricacy of modern financial crimes. Oracle's latest innovation directly addresses this challenge, ushering in a new era of AI-driven investigations.

A Paradigm Shift in Financial Crime Investigations

"The addition of agentic AI capabilities to our Investigation Hub Cloud Service represents a paradigm shift in financial crime investigations,"¹ stated Jason Somrak, head of financial crime product strategy at Oracle Financial Services. "Our unique generative AI approach follows investigative plans, collects evidence, and recommends actions while providing investigators with robust narratives documenting the findings. This enables firms to drive consistency in decision-making and thoroughly investigate all risks automatically while realizing massive operational efficiencies."²

The core of this enhancement lies in the deployment of a broad class of AI agents, each designed to perform specific investigative tasks. These agents are not merely chatbots that respond to user queries; they are proactive, intelligent systems capable of:

Automated Data Collection: AI agents can automatically gather and synthesize data from diverse sources, including internal databases, external watchlists, and public records, significantly reducing the time spent on manual data retrieval.
Pattern Recognition and Analysis: Leveraging advanced machine learning algorithms, these agents can identify subtle patterns and anomalies that may indicate financial crime, even in vast datasets.
Narrative Generation: A key innovation is the agents' ability to generate comprehensive, human-readable narratives that summarize their findings. These narratives provide investigators with clear and concise reports, facilitating faster and more informed decision-making.
Risk Assessment and Prioritization: AI agents can assess the risk associated with each case, enabling investigators to prioritize their efforts and focus on the most critical leads.
Agentic Workflows: The agents work together in orchestrated workflows, allowing for complex investigative tasks to be automated.

Addressing the Challenges of Modern Financial Crime

Financial institutions are under immense pressure to comply with stringent regulatory requirements while simultaneously combating the ever-evolving tactics of criminals. Traditional investigative processes are often:

Time-Consuming: Manual data gathering and analysis can take days or even weeks, delaying critical investigations.
Resource-Intensive: Financial institutions must dedicate significant personnel resources to investigative tasks.
Prone to Human Error: Manual processes are susceptible to errors and inconsistencies, which can compromise the accuracy of investigations.
Difficulty with complex pattern recognition: Humans can have difficulty seeing the patterns that AI can quickly see.

Oracle's AI agents address these challenges by automating key aspects of the investigative process, freeing up human investigators to focus on higher-level analysis and decision-making.

Key Benefits for Financial Institutions

The integration of AI agents into the Investigation Hub Cloud Service offers numerous benefits for financial institutions, including:

Increased Efficiency: Automation of data collection and analysis significantly reduces investigation times.
Improved Accuracy: AI-driven pattern recognition and analysis enhance the accuracy of investigations.
Reduced Operational Costs: Automation reduces the need for manual labor, leading to cost savings.
Enhanced Compliance: Improved investigative capabilities help financial institutions meet regulatory requirements.
Faster Detection of Financial Crime: The ability to quickly identify and investigate suspicious activity enables financial institutions to mitigate risks and prevent losses.
Enhanced consistency: AI driven processes create more consistent results.

Generative AI and the Future of Investigations

The use of generative AI to create investigative narratives is a particularly noteworthy advancement. This capability transforms raw data into actionable insights, providing investigators with a clear and comprehensive understanding of each case.

Oracle's commitment to innovation in financial crime prevention is evident in its continuous development of advanced AI-powered solutions. The company's focus on integrating generative AI and agentic workflows reflects a broader trend toward the automation of complex tasks in the financial sector.

Global Availability and Impact

The enhanced Investigation Hub Cloud Service is now available globally to financial institutions of all sizes. This accessibility ensures that organizations worldwide can leverage the power of AI to strengthen their financial crime prevention efforts.

The impact of this technology is expected to be significant, as financial institutions strive to keep pace with the ever-evolving landscape of financial crime. By automating key investigative processes, Oracle's AI agents empower organizations to detect and prevent financial crime more effectively, contributing to a safer and more secure financial system.

In conclusion, Oracle Financial Services' integration of AI agents into its Investigation Hub Cloud Service represents a major advancement in financial crime prevention. By automating complex investigative processes, these AI agents empower financial institutions to detect and prevent financial crime more effectively, contributing to a safer and more secure financial system.

*1 https://www.oracle.com/news/announcement/oracle-brings-ai-agents-to-the-fight-against-financial-crime-2025-03-13/

*2 https://www.oracle.com/news/announcement/oracle-brings-ai-agents-to-the-fight-against-financial-crime-2025-03-13/

DEEPSEEK SPOOKED THE MARKET, THIS AI STOCK CASHED THE CHECK

Mad Hedge AI

(AVGO), (MSFT), (AMZN), (GOOG), (META), (INTC), (MRVL)

The panicked messages flooded my inbox last week as tech stocks took a nosedive.

One subject line simply read: "BROADCOM - SELL NOW???" Another longtime reader called me at 5:30 AM (clearly forgetting about time zones) asking if the AI bubble had finally burst. His portfolio was bleeding red and he was ready to hit the eject button on everything tech-related.

It's times like these when having decades of market experience comes in handy. I've seen this movie before - different characters, same plot.

While panicked investors were selling, Broadcom delivered Q1'25 earnings that sent shares up by over 8%. Despite Microsoft (MSFT) pulling some data center leases, the recent selling has created a significant buying opportunity for AVGO shares.

The numbers tell the real story here. Broadcom's semiconductor business grew 11% year-over-year, with 50% of the $8.2 billion in semiconductor revenue coming from AI infrastructure.

When half your semiconductor business is driven by the hottest sector in tech, you've positioned yourself exactly where you need to be.

Why such strong AI demand? Sources at Microsoft, Amazon (AMZN), and Google (GOOG) confirm the DeepSeek V3 models have lit a fire under their AI development teams.

This has accelerated the timeline for all major players to improve their AI systems, with hyperscalers now racing to deploy more advanced models.

Broadcom's management isn't sitting idle amid this AI acceleration. They're ramping R&D investments in next-generation accelerators, targeting 2nm AI XPUs with 10,000 teraflops capability.

For context, that's massive computational power that will drive the next wave of AI applications. They're also enabling clusters scaling to 1 million XPUs - the infrastructure backbone needed for training tomorrow's frontier models.

The technical progress is equally impressive.

A senior Broadcom engineer I met last month revealed they've doubled RAID capacity for existing Tomahawk sites and will begin sampling 1.6T bandwidth switches soon.

These advances put Broadcom at the center of the AI infrastructure buildout that every major tech company needs.

All this positions Broadcom for explosive growth.

By FY27, their AI addressable market is projected to reach $60-90 billion. They've already added two more hyperscalers as customers, bringing the total to four.

The company now develops custom AI chips for Alphabet, Microsoft, Amazon, Meta (META), and ByteDance (the parent company of TikTok) - essentially becoming the arms dealer in the AI arms race.

The near-term outlook is equally promising. For Q2'25, Broadcom expects AI revenue to reach $4.4 billion, up 44% from last year. These aren't small percentage gains on tiny revenue bases - we're talking billions in growth from already substantial numbers.

Beyond AI, the picture is more nuanced but still positive overall.

Traditional semiconductors face some growth pressures, though telecom customers should increase spending in Q2'25. Enterprise networking will likely remain flat in the first half of the year due to inventory digestion.

Meanwhile, enterprise software has been a bright spot, growing 47% year-over-year.

Management is successfully converting perpetual licenses to subscription models, with 60% converted so far.

This transformation to recurring revenue improves visibility and stability in the business model - something Wall Street consistently rewards with higher multiples.

Innovation continues across the business.

One notable example is their VMware Private AI Foundation with NVIDIA, already adopted by 39 enterprise customers.

This platform helps companies manage GPU infrastructure for their AI workloads - positioned exactly where enterprise spending is heading as AI moves from experimental to production environments.

Looking at the financials, the picture remains strong. For Q2'25, I'm forecasting $14.9 billion in revenue and $1.59 EPS.

Inventory levels increased to $1.9 billion (56 days on hand) to support the AI production ramp-up.

The company has managed its balance sheet well, paying down $495 million in fixed-rate and $7.6 billion in floating-rate debt, ending with a reasonable 1.64x leverage ratio.

Is there risk? Certainly.

If Microsoft's canceled data center leases truly signal excess compute capacity across the industry, we could see a broader slowdown in XPU investments.

But the capital expenditure guidance from all major hyperscalers suggests this is isolated, not systemic.

Valuation remains attractive despite the growth story.

AVGO shares currently trade at a significant discount to peers like Marvell Technology (MRVL) in the custom ASIC market.

In fact, Broadcom CEO Hock Tan explicitly stated they're "too busy" with AI and VMware to consider an Intel (INTC) purchase - a signal of management's focus on organic growth opportunities rather than distracting acquisitions.

Looking at the broader market context, the last two years delivered remarkable gains with minimal pullbacks. We're now experiencing that needed pullback, which could continue for 3-6 months.

After that, I expect markets to finish 2025 in positive territory, with AI leaders like Broadcom leading the charge.

My response to those panicked emails was simple: use this opportunity to build a position in AVGO.

When a company dominates its space with 50% of revenue coming from the fastest growing segment in tech, you don't sell - you buy more.

And yes, in case you're wondering, I own the stock. I've been adding on every dip.

Together, AI: Democratizing Access to Foundation Models, One API Call at a Time

Mad Hedge AI

In the burgeoning landscape of artificial intelligence, where massive foundation models (FMs) reign supreme, a new player has emerged, promising to democratize access and accelerate innovation. Together, AI, a startup founded by a team of seasoned AI researchers and engineers, is rapidly gaining traction with its platform designed to simplify the development and deployment of cutting-edge AI models.

At the heart of Together, AI's mission lies a fundamental belief: the power of foundation models should not be confined to a select few with access to vast computational resources. Instead, they envision a future where developers, researchers, and businesses of all sizes can leverage these transformative technologies to build groundbreaking applications.

Bridging the Gap: Simplifying Foundation Model Access

The challenge with FMs, like large language models (LLMs) and diffusion models, is their sheer size and complexity. Training and deploying these models requires significant computational power, specialized expertise, and substantial financial investment, creating a barrier for many. Together, AI aims to dismantle this barrier by offering a unified platform that simplifies the entire lifecycle of working with FMs.

Their platform provides a comprehensive suite of tools and services, including:

API-driven access to a diverse range of FMs: Together, AI offers a curated selection of state-of-the-art models, including their own open-source models as well as those from leading research labs, accessible through a simple API. This eliminates the need for developers to manage complex infrastructure or navigate the intricacies of model deployment.
Optimized infrastructure for efficient model execution: The platform leverages a distributed infrastructure optimized for running large-scale AI models, ensuring fast inference speeds and low latency. This allows developers to focus on building applications rather than worrying about infrastructure management.
Tools for fine-tuning and customizing models: Together, AI recognizes that pre-trained models often need to be adapted to specific tasks and domains. Their platform provides intuitive tools for fine-tuning models with custom datasets, enabling developers to tailor the models to their unique requirements.
Open source contributions and community focus: Together, AI is deeply committed to open source and actively contributes to the development of open foundation models. They believe in fostering a collaborative community where researchers and developers can share knowledge and contribute to the advancement of AI.

The Power of Open Source: RedPajama and Beyond

One of Together, AI's most notable contributions is the RedPajama project, an initiative to create a fully open-source reproduction of the LLaMA language model. This project exemplifies their commitment to transparency and accessibility in AI.

RedPajama aims to address the concerns surrounding the proprietary nature of many leading LLMs, providing a freely available alternative that can be used for research and development without restrictions. By democratizing access to powerful language models, RedPajama has the potential to accelerate innovation and foster a more open and inclusive AI ecosystem.

Beyond RedPajama, Together, AI is actively involved in other open-source initiatives, contributing to the development of tools and libraries that simplify the use of foundation models. They believe that open collaboration is essential for driving progress in AI and ensuring that the benefits of this technology are widely shared.

Use Cases and Applications: Transforming Industries

The impact of Together, AI's platform is already being felt across a wide range of industries. Developers are using their tools to build innovative applications in areas such as:

Natural Language Processing (NLP): Building chatbots, text summarization tools, and content generation systems with unprecedented accuracy and fluency.
Computer Vision: Developing image and video analysis applications, including object detection, image generation, and video editing tools.
Drug Discovery and Materials Science: Accelerating research by using FMs to analyze complex datasets and predict the properties of new molecules and materials.
Software Development: Enhancing developer productivity with AI-powered code generation and debugging tools.
Personalized Education: Creating adaptive learning platforms that tailor educational content to individual student needs.

For instance, a startup developing a customer service chatbot can leverage Together, AI's platform to access a powerful LLM, fine-tune it with their customer support data, and deploy it quickly and easily. Similarly, a research team working on drug discovery can use the platform to analyze vast datasets of chemical compounds and identify potential drug candidates.

The Future of Foundation Models: A Collaborative Vision

Together, AI's vision extends beyond simply providing access to existing models. They are actively working on developing new and more efficient FMs, pushing the boundaries of what is possible with AI.

Their research efforts focus on areas such as:

Efficient model training and inference: Developing techniques to reduce the computational cost of training and deploying FMs, making them more accessible to a wider range of users.
Multimodal models: Building models that can process and integrate information from multiple modalities, such as text, images, and audio, enabling more sophisticated and versatile applications.
Responsible AI: Developing tools and techniques to mitigate the risks associated with FMs, such as bias and misinformation, ensuring that these technologies are used ethically and responsibly.
Decentralized AI: Exploring the potential of decentralized approaches to training and deploying FMs, creating more robust and resilient AI systems.

They emphasize the importance of collaboration and believe that the future of AI lies in a decentralized and open ecosystem. This vision is reflected in their commitment to open source and their active engagement with the AI research community.

Challenges and Considerations: Navigating the AI Landscape

While Together, AI's platform offers significant advantages, it is important to acknowledge the challenges and considerations associated with the use of foundation models.

Bias and fairness: FMs can inherit and amplify biases present in their training data, leading to unfair or discriminatory outcomes. Addressing these biases requires careful data curation, model evaluation, and ongoing monitoring.
Misinformation and misuse: The ability of LLMs to generate realistic and persuasive text raises concerns about the potential for misuse, such as generating fake news or impersonating individuals. Robust safety measures and ethical guidelines are essential to mitigate these risks.
Computational cost: Even with optimized infrastructure, running large FMs can be computationally expensive. Finding ways to reduce the cost of inference and training is crucial for making these technologies more accessible.
Data privacy: Training FMs often requires access to large amounts of data, raising concerns about data privacy. Developing privacy-preserving techniques is essential for ensuring that data is used responsibly.

Together, AI is actively addressing these challenges through its research efforts and its commitment to responsible AI practices. They believe that by working together, the AI community can overcome these obstacles and ensure that the benefits of FMs are realized for all.

The Rise of a New AI Paradigm

Together, AI represents a new paradigm in the development and deployment of foundation models. By democratizing access and fostering collaboration, they are empowering developers, researchers, and businesses to build groundbreaking applications that were previously unimaginable.

Their commitment to open source, their focus on efficient infrastructure, and their dedication to responsible AI practices are positioning them as a key player in the rapidly evolving AI landscape. As foundation models continue to advance and become more widely accessible, platforms like Together, AI will play a crucial role in shaping the future of artificial intelligence.

Deaglo: Revolutionizing FX Risk Management with AI

Mad Hedge AI

In the dynamic world of international finance, where currency fluctuations can significantly impact business outcomes, effective foreign exchange (FX) risk management is paramount. Deaglo, a pioneering FinTech company, is at the forefront of this critical domain, leveraging cutting-edge technology to empower businesses and financial institutions with unprecedented control over their FX exposures.

A Visionary Approach to FX Risk:

Deaglo's mission is to revolutionize the way businesses navigate the complexities of the global FX market. Recognizing the limitations of traditional, often manual, FX risk management processes, Deaglo embarked on a journey to develop innovative solutions that harness the power of artificial intelligence (AI) and advanced analytics.

The FX Assistant: An AI-Powered Game-Changer:

A cornerstone of Deaglo's innovation is the FX Assistant, a groundbreaking AI-driven platform designed to streamline and optimize FX risk management for a wide range of clients, from multinational corporations to investment banks.

Key Features and Benefits:

Real-time Insights: The FX Assistant provides real-time market data and analytics, enabling clients to make informed decisions based on the latest market trends.
Automated Trade Recommendations: Leveraging sophisticated AI algorithms, the platform generates personalized trade recommendations, optimizing hedging strategies and minimizing potential losses.
Enhanced Efficiency: By automating many of the time-consuming tasks associated with FX risk management, the FX Assistant frees up valuable time for finance professionals to focus on strategic initiatives.
Improved Decision-Making: The platform provides actionable insights and visualizations, empowering users to make more informed and confident decisions regarding their FX exposures.
Seamless Integration: The FX Assistant seamlessly integrates with existing trading platforms and systems, ensuring a smooth and efficient workflow.

Impact on the Financial Landscape:

The impact of Deaglo's innovations is already being felt across the financial landscape. By providing businesses with the tools and insights they need to effectively manage their FX risk, Deaglo is:

Reducing Volatility: By optimizing hedging strategies and mitigating currency fluctuations, Deaglo helps businesses protect their bottom line and improve financial stability.
Increasing Efficiency: Automation and streamlined workflows lead to significant time and cost savings for businesses of all sizes.
Improving Decision-Making: Access to real-time data and AI-powered insights empowers businesses to make more informed and strategic decisions regarding their FX exposures.
Driving Innovation: Deaglo's pioneering approach is driving innovation in the FinTech space, pushing the boundaries of what is possible in FX risk management.

A Commitment to Excellence:

Deaglo's success is built on a foundation of unwavering commitment to excellence. The company boasts a team of highly skilled professionals with deep expertise in finance, technology, and risk management. This dedicated team is constantly striving to improve the FX Assistant and develop new solutions that address the evolving needs of the market.

Looking Ahead:

As the global economy continues to evolve and become increasingly interconnected, the importance of effective FX risk management will only grow. Deaglo is well-positioned to remain at the forefront of this critical domain, leveraging its innovative technology and unwavering commitment to customer success to shape the future of FX risk management.

In Conclusion:

Deaglo is not just another FinTech company; it is a visionary organization that is transforming the way businesses approach FX risk management. By harnessing the power of AI and advanced analytics, Deaglo is empowering clients to navigate the complexities of the global FX market with confidence and control. As the company continues to innovate and expand its offerings, it is poised to play an even greater role in shaping the future of international finance.

AN AI PLAYER NOBODY'S TALKING ABOUT

Mad Hedge AI

(CLS), (MSFT), (GOOG), (GOOGL), (NVDA)

I was driving back from Salt Lake City a few weeks ago when I got a call from a former Concierge Service client who runs a drone manufacturing business.

His story is remarkable - what started as a hobby five years ago now generates $40 million in revenue, with military contracts on the horizon promising even more growth.

He keeps telling me that our weekly calls during those crucial early days were what kept him from selling to the first European conglomerate that waved a checkbook in his face.

"You told me to hold out for ten times what they were offering," he reminded me. I don't remember saying that, but who knows?

After thousands of client calls over the years, they tend to blur together. Still, he swears it was my portfolio review that convinced him to plow every nickel of profit back into engineering when his competitors were cashing out.

I'm not about to take credit for his success – the guy's a genius in his own right – but I'd be lying if I said it didn't make my day to hear that. Now he's being bombarded with takeover offers from European and Asian firms desperate for new profit streams at any cost.

That conversation got me thinking about AI stocks and where the real opportunities might be hiding.

AI fever is alive and well on Wall Street. But while tech giants like Alphabet (GOOG) (GOOGL) and Microsoft (MSFT) operate under the shadow of China's DeepSeek, the architects of artificial intelligence systems—from semiconductors to servers to sprawling data centers—have emerged as the winners so far in 2025.

Look no further than Nvidia (NVDA), which saw its sales and profits surge as tech companies threw billions at its advanced chips.

The demand for its Blackwell series alone generated a staggering $11 billion in revenue last quarter. But with its valuation soaring, NVDA is no longer the under-the-radar opportunity it once was.

While Nvidia's dominance is undeniable, the real money in AI isn’t just in chips—it’s in the infrastructure that keeps the entire ecosystem running.

And that’s where Celestica (CLS) comes in. While most investors were fixated on Nvidia, CLS quietly delivered massive gains, positioning itself as a crucial player in AI’s supply chain.

Celestica makes the electronic guts that keep AI data centers running. It’s not as flashy as ChatGPT or robotics breakthroughs, but the companies that build and maintain AI infrastructure are often where the real money is made. It’s the same reason drone component suppliers have been making a fortune while the spotlight stays on the drones themselves.

And the numbers back it up. Celestica reported $2.55 billion in Q4 revenue, an adjusted EPS of $1.11, and a forecast of $10.7 billion in revenue for 2025—a 22% earnings growth rate in a market where consistent growth is getting harder to find.

Yet its valuation remains reasonable, with a PEG ratio of 0.8, meaning investors are paying less per unit of growth compared to sector averages.

In 2025, market volatility has been brutal. Geopolitical tensions, persistent inflation, and Donald Trump’s new tariffs on China helped fuel a 5% Nasdaq drop last week, sending investors into panic mode.

But CLS kept climbing. Over the last three months, while the broader market wobbled, Celestica delivered a nearly 29% gain.

Even better, nine analysts have revised their earnings estimates upward in the last 90 days, with zero downward revisions. That’s the kind of confirmation bias I can get behind.

It’s a strategy I’ve seen pay off before. My former client reinvested in engineering instead of selling out early, betting on long-term value over a quick exit.

Celestica has been following a similar playbook—focusing on becoming an indispensable part of the AI infrastructure rather than competing with the companies making headlines.

While others chase the next big AI breakthrough, Celestica is already supplying the backbone that powers them all.

It currently ranks as the top electronic manufacturing services stock and is among the top 10 technology stocks overall—solid proof to its growing influence in the industry.

And with hyperscaler demand accelerating and AI adoption still in its early innings, Celestica is positioned to become even more critical to the AI revolution in the coming years.

So, where is the next 10-bagger in AI? It's probably not limited to the household names that dominate the headlines.

As I've seen time and again throughout my career, sometimes the most profitable investments are found in the companies building the essential tools for the gold rush, not the miners themselves.

And just like my drone-building client who turned personalized investment advice into a $40 million business by avoiding the quick exit, the real winners in AI will be those who recognize the opportunities hiding in plain sight.

After all, fortunes aren’t made by chasing shiny objects—they’re made by supplying the circuit boards that keep them running.

THOSE AREN'T CHICKENS IN YOUR TRANSLATOR APP

Mad Hedge AI

AI Tech Letter

(THOSE AREN'T CHICKENS IN YOUR TRANSLATOR APP)

(IBEX)

I was flying back from Monterey last month when I found myself eavesdropping on a fascinating exchange at the customer service counter.

The passenger ahead of me—clearly not a native English speaker—was struggling to explain his situation to an agent who spoke only English.

Then, the agent tapped her screen, and suddenly, she was responding in his language with surprising accuracy.

Curious, I leaned in as she handed him a feedback form. "How was your experience with our new AI-powered translation service?" it read.

When my turn came, she confirmed the system was new. "It’s a game-changer," she said. "Three weeks ago, we'd have needed a translator or spent half an hour playing charades."

This experience came to mind as I dug into IBEX Limited (IBEX), a Washington D.C.-based company positioning itself at the intersection of old-school customer service and cutting-edge AI.

Their business process outsourcing services have earned them a reputation for high-quality omnichannel support across growing offshore markets, with a focus on using AI, automation, and data analytics to transform how brands engage with customers.

Remember when "customer service" meant being stuck on hold for an hour while a recorded voice assured you that your call was "very important"? IBEX is changing that narrative.

Their Wave iX solutions don't replace humans—they enhance them. One hospitality client now deploys English-speaking agents with AI-powered translation to handle multilingual interactions (goodbye language barriers and soul-crushing hold music).

The quality of these translations is what sets them apart in a world where most translation apps still turn phrases like "welcome gift" into "live chickens" (true story from my last international trip).

Their AI also handles routine inquiries ("What are your hours?") while humans tackle the complex ones ("Why does my new dishwasher sound like it's harboring raccoons?").

It's essentially digital triage at its finest—at least until that gets disrupted too, which in Silicon Valley terms means sometime next Tuesday.

Now, let's talk numbers. IBEX posted a record-breaking second quarter with revenues hitting $140.7 million, up 6.1% year-over-year.

Their "land-and-expand playbook" brought in five new major clients this quarter alone. Earnings per share came in at $0.59 (beating expectations by $0.08), with adjusted EBITDA climbing to $16.5 million and margins improving to 11.8%.

If you're new to investing, beating analyst expectations is like showing up to your in-laws' house with an expensive bottle of wine—it won't solve all your problems, but it certainly helps set a positive tone.

The sector breakdown tells another interesting story.

Offshore operations (now 53% of total revenue) grew by 14%, while HealthTech surged 31%—no surprise to anyone who's visited a doctor's office lately and watched them wrestle with patient management software.

Travel, logistics, and retail also saw healthy increases. A significant move this quarter was IBEX's buyback of 3.6 million shares from TRGI, freeing it from "controlled company" status and helping boost EPS by 36%.

But let's not pop the champagne just yet. For all its strong earnings, IBEX has some glaring weak spots.

Revenue growth is anemic at 1.66%—well below the sector median of 3.93%. The FinTech segment is particularly troubling, with revenue down 15%.

Cash flow remains negative (they burned through $3.2 million), and their net cash position dropped precipitously from $60.8 million to just $13.7 million thanks to that $70 million share buyback.

It's like bragging that you're only losing $100 at the poker table instead of $200—technically an improvement, but your wallet still gets lighter.

Management has also acknowledged that costs related to AI implementation and infrastructure are rising, potentially squeezing margins in the latter half of fiscal 2025.

Meanwhile, customers are taking longer to pay (Days Sales Outstanding extended from 75 to 79 days), which could further strain working capital.

So what's the bottom line? At its current valuation—9.97x forward P/E for 2025 and 8.61x for 2026—IBEX might represent a bargain for investors who believe in the long-term promise of AI-driven outsourcing.

I've been trading long enough to know that sometimes the market offers gifts for those willing to look past short-term turbulence.

For the stock to realize its potential upside, however, IBEX needs to demonstrate stronger revenue growth and better cash flow management.

If it succeeds, a move toward a 12x–14x P/E range would offer meaningful upside. If not, it could remain stuck in what I call "value trap purgatory."

As for me, I'm keeping IBEX on my watchlist—right next to “learn a second language” and “never fly standby again.” Some opportunities, like some flights, are worth waiting for.