01.31.2025

ForwardThinking

The whiplash-inducing pace of change

This week it would have been easy to write my ForwardThinking commentary about DeepSeek, the new model from China that had everyone buzzing about its impact and speculating what it means. Find your favorite writer or thought leader and there was surely an opinion. But what struck me about the DeepSeek news wasn’t just the weightiness of it, although it clearly was a bombshell, it was its part in the breakneck news cycle we are experiencing.

Consider that in the last week we’ve heard about OpenAI releasing a beta of Operator, a new AI agent that reportedly can search the web and undertake tasks on your behalf. Then came Stargate, a project to build massive data centers for OpenAI with an astonishing $500 billion potential price tag, a number that’s so big, it’s hard to wrap our heads around. Then DeepSeek came along to blow our minds yet again.

DeepSeek in particular raised a plethora of questions about everything we’ve been told about generative AI. If this model, which seems quite capable, can be produced for millions of dollars, why do U.S. tech companies require hundreds of billions to build out their data centers in the chase for ever more compute power and presumably increasingly sophisticated and capable models? What does it mean for the cost of AI in the future, and are we witnessing the beginning of the commoditization of generative AI?

That’s just one week of news folks. It’s not just you. It is hard to process what it all means. I do this for a living, and it’s hard for me. I’m sure it’s even more challenging for anyone trying to set tech policy inside a large company with what is essentially a moving target.

It feels like we’re constantly falling behind, and we need to keep accelerating to keep up. Yet we know that large organizations don’t tend to move fast, and in spite of the high-speed pace of the announcements coming from vendors, study after study is showing that very few companies are actually implementing successful generative AI projects at scale. In fact, most are still very much in the experimenting phase.

A recent report from Deloitte, The State of Generative AI in the Enterprise concluded that it’s going to take time to scale this technology and executives need to be patient. “The GenAI journey is long, and C-suite leaders need to be realistic about time horizons for project success and organizational transformation,” the report found.

It’s easy to feel like you’re lagging when one big news event after another comes washing over us like so many waves upon the shore, but the fact is we are very early with this technology. The vendors may be pushing the velocity, but enterprises are going to do what enterprises do – move slowly and deliberately and figure out what works and what doesn’t.

So slow down and take a deep breath because we’re just getting started, and we are going to have to work hard to separate the hype from the reality.

- Ron

What's new on FastForward

Opinions come pouring in on $500 billion Stargate Project, as DeepSeek throws a curve

As I wrote in this week's ForwardThinking commentary, the tech news cycle has been moving with breathtaking speed, and if you needed proof, last week's whiplash-inducing announcements around the $500 billion Stargate Project, quickly followed by DeepSeek's release of a newer, cheaper model has left the industry wondering exactly what's coming next, and what it all means.

I spoke to some industry experts and included commentary from leading journalists and commentators to try and make sense of the news.

Read the full article

Intuit CTO puts his staff of 8000 engineers in charge of modernizing the tech stack

As CTO at a company with 8000 engineers and a market cap of almost $170 billion, Alex Balazs has a lot of responsibility. And fighting tech debt in a company that's over 40 years old is part of that. He says one of the strategies he employs to keep the tech stack from getting stale is getting that large team of engineers involved.

"For some of our bottoms up initiatives our engineers are encouraged to spend some percentage of their time basically improving the existing technology stack, rewriting old, crufty code, updating frameworks, whatever the case may be," he said.

Read the full article

Databricks CIO says it's important to align substantive changes with the needs of the business

While technically still a startup, Databricks has the size and scope of a much more mature company, and that takes a special kind of technology leadership.

When it comes to implementing new technologies like generative AI, Databricks CIO Naveen Zutshi says that it's important to align the implementation with the needs of the business units, getting them involved to build more practical solutions.

"Folks are used to doing things a certain way, and for you to change your behavior, to do something differently, is a big shift.” He tries to manage that by incorporating those tech changes into the flow of business, and that can help ease that anxiety that employees may have about introducing new ways of doing things.

Read the full article

In spite of reports projecting massive job loss from AI, the future could involve deeper human-machine connections

You couldn't help but notice the screaming headlines this week projecting massive job loss from AI, but as I dug deeper into the reports, I found that the picture wasn't quite as grim as the headlines might suggest, and it's possible that humans will find ways to adjust to this new reality.

Read the full article

News Bytes

DeepSeek release Wreaked Havoc in Silicon Valley this week

I would be remiss if I didn't write something about DeepSeek in this week's news section. To recap briefly, DeepSeek is a new AI model from a Chinese startup that reportedly was able to train a highly effective model for a fraction of the cost of OpenAI and other generative AI models on the market.

The week opened with news of its release, which sent shockwaves from Silicon Valley all the way to Wall Street. Consider that the news triggered an astonishing $600 billion loss in market cap for GPU chip maker Nvidia, which according to CNBC, was the largest one-day loss in the history of the market. (Weep not for Nvidia, however, whose market cap still sits at 3.05 trillion as of this morning.)

Just about everyone had an opinion about whether this was a positive advancement or a serious threat to the biggest hyperscaler tech companies, to national security, to U.S. AI dominance, to you name it. It didn't take long for people to suggest that maybe DeepSeek wasn't what it seemed. OpenAI, a company that could rightly be threatened by DeepSeek's cheaper approach to model building, led the charge accusing the company of stealing its data to train its models. It's worth noting that many found the accusation ironic given that OpenAI itself hoovered up data, often without permission, to train its models.

However this plays out, DeepSeek has raised some interesting questions about the cost of training models and the future of generative AI. Even though it isn't clear whether their cost claims are true or not, the company's capable model, one that comes from China, clearly had people questioning their conventional wisdom about generative AI.

Justice department looks to block $14 billion Hpe-Juniper Networks deal

Tech companies that were hoping that a change in administration would ease regulatory oversight of large acquisitions might have been surprised to learn yesterday that the Trump Justice Department is suing to block the $14 billion deal between HPE and Juniper Networks. The deal was announced almost exactly a year ago, and has been in oversight limbo ever since.

At the time, both companies claimed it was about synergy around network and edge computing, especially as it related to AI (because everything these days is related to AI). In a blog post announcing the deal, Juniper CEO Rami Rahim wrote, “This combination with HPE is expected to enable us to deliver more comprehensive, more competitive, truly end-to-end experience-first AI-native solutions.”

The DoJ apparently isn't buying that argument. In a press release announcing the lawsuit, it stated that "the proposed transaction would eliminate fierce head-to-head competition between the companies, raise prices, reduce innovation, and diminish choice for scores of American businesses and institutions, in violation of Section 7 of the Clayton Act."

The two companies, as you can imagine, vehemently disagreed. "We believe the Department of Justice’s analysis of this acquisition is fundamentally flawed and we are disappointed in its decision to file a suit attempting to prohibit the closing of the transaction," the companies wrote in a statement.

The DoJ lawsuit is even more surprising, given that the EU and the UK have already approved the deal. Both tend to scrutinize deals carefully through a competitive lens, and both still approved the deal without conditions

Microsoft cloud business slows

Microsoft reported earnings this week, and while its Azure cloud business still grew a brisk 31% YoY, it was down from 33% over the prior quarter. While that news didn't exactly send investors into a panic, it did raise concerns, especially since as Bloomberg reports, CFO Amy Hood talked about capacity constraints in data centers. In other words, they don't have enough data centers to meet demand.

That would seem to be a good problem to have, assuming they can build data centers fast enough, and customers don't flee to competitors in the interim. Hood suggested in comments to investors that they should be able to catch up at the end of the year, but the fact is that the company's shift to AI is using up vast amounts of compute resources, something that just about every company hoping to take advantage of the generative AI boom is finding. But there are limits to capacity, even for a company as large as Microsoft, and it seems for now at least, they can't build data centers fast enough for today's requirements.

U.S. copyright office rules ai-generated content can be copyrighted, so long as a human is involved

In an interesting ruling this week, the U.S. Copyright office, which typically seeks to protect work like books, movies and TV shows made by humans, ruled that in some cases, it will allow copyrights for AI-generated works. But there was an important caveat that humans have to be involved in a substantial way. Works created strictly by AI would not be subject to copyright protection.

The office ruled that like other copyright applications, the decision to give a copyright to AI-assisted works will be determined on a case-by-case basis, but as AI becomes increasingly capable, it will be much more challenging for humans to make the distinction between AI-generated works and human-AI generated works.

What we're reading

British Museum Reading Room with book shelves lining the walls and desks for reading in the center of the room.

The British Museum Reading Room. CC BY-SA 4.0 via Wikimedia Commons

404 Media

OpenAI Furious DeepSeek Might Have Stolen All the Data OpenAI Stole From Us

Jason Koebler

Generative AI ROI: Don’t Forget to Invest in Your People

Christina Stathopoulos

TechCrunch

Climate change ignited LA’s wildfire risk — these startups want to extinguish it

Tim De Chant

Cybersecurity Dive

UnitedHealth hikes number of Change cyberattack breach victims to 190M

Emily Olsen

Look who's talking

“Simply put, LLM margin compression is here. The scary part is the LLM giants didn't have profit margins to begin with. LLMs are going to commodity in a hurry.”

- Larry Dignan, Constellation Insights Blog

Please subscribe to the newsletter and feel free to share.

3350 Virginia St, Suite 262, Miami, FL 33133

ForwardThinking

The whiplash-inducing pace of change

What's new on FastForward

Opinions come pouring in on $500 billion Stargate Project, as DeepSeek throws a curve

Intuit CTO puts his staff of 8000 engineers in charge of modernizing the tech stack

Databricks CIO says it's important to align substantive changes with the needs of the business

In spite of reports projecting massive job loss from AI, the future could involve deeper human-machine connections

News Bytes

DeepSeek release Wreaked Havoc in Silicon Valley this week

Justice department looks to block $14 billion Hpe-Juniper Networks deal

Microsoft cloud business slows

U.S. copyright office rules ai-generated content can be copyrighted, so long as a human is involved

What we're reading

404 Media

OpenAI Furious DeepSeek Might Have Stolen All the Data OpenAI Stole From Us

LinkedIn

Generative AI ROI: Don’t Forget to Invest in Your People

TechCrunch

Climate change ignited LA’s wildfire risk — these startups want to extinguish it

Cybersecurity Dive

UnitedHealth hikes number of Change cyberattack breach victims to 190M

Look who's talking

Subscribe

FastForward #7: The whiplash-inducing pace of change

You may also like...