Introduction to MLPerf Benchmarks

As you AI pros know, the 125-member MLCommons organization alternates training and inference benchmarks every three months. This time around, its all about training, which remains the largest AI hardware market, although not by much as inference drives more growth as the industry shift from research (building) to production (using). As usual, Nvidia took home all the top honors.

AMD Joins the Training Party

For the first time, AMD joined the training party (they had previously submitted inference benchmarks), while Nvidia trotted out their first GB200 NVL72 runs to demonstrate industry leadership. Each company focussed on their best features. For AMD it is larger HBM memory, while Nvidia exploited its Arm/GPU GB200 superchip and NVLink scaling.

The Bottom Line

The bottom line is that AMD can now compete head to head with H200 for smaller models that fit into MI325’s memory. That means AMD cannot compete with Blackwell today, and certainly cannot compete with NVLink-enabled configurations like NVL72.

AMD: Its All About The Memory

AMD has more HBM memory on their MI325 platform than any Nvidia’s GPU, and can therefore contain an entire medium-sized model on a single chip. So, they ran the training benchmark that fits, the Llama 2-70B LORA model. The results are reasonably impressive, besting the Nvidia H200 by an average of 8%. While a good result, I doubt many would choose AMD for 8% better performance, even at a somewhat lower price. The real question, of course, is how much better the MI350 will be when it launches next week, likely with higher performance and even more memory.

AMD’s Limitations

One thing AMD will not offer soon is better networking for scale-up; the UA-Link needed to compete with NVLink is still many months away (possibly in the MI400 timeframe in 2026). So, if you only need a 70B model, AMD may be a better deal than Nvidia H200; but not by much.

Traction with Partners

AMD is also showing traction with partners, and better performance from its ROCm software, which took quite a beating from SemiAnalysis last December. With better ease-of-use from ROCm, partners can benefit from offering customers a choice; many enterprises do not need the power of an NVL72 or NVLink, especially if they are focussed on simple inference processing. And of course, AMD can offer better availability, as NVIDIA GB200 is much harder to obtain due to overwhelming demand and pre-sold capacity. The rumor mill says GB200 still takes over a full year delivery time if you order today.

Nvidia: Its All About Scale-Up

Nvidia says the GB200 NVL72 has now arrived, if you were smart enough to put in an early order. With over fifty benchmark submissions using up to nearly 2500 GPUs, Nvidia and their partners ran every MLPerf benchmark on the ~3000 pound rack, winning each one. CoreWeave submitted the largest configuration, with nearly 2500 GPUs.

Nvidia’s Advantage

While the GB200 NVL72 can outperform Hopper by some 30X for inference processing, its advantage for training is “only” about 2.5X; thats still a lot of savings in time and money. The reason is that inference processing benefits greatly from the lower 4- and 8-bit precision math available in Blackwell, and the new Dynamo "AI Factory OS” optimizes inference processing and reuses previously calculated tokens in KV-Cache.

My Takeaway

While AMD does not yet have the scale-up networking required to train larger models at Nvidia’s level of performance, this benchmark shows that they are getting close enough to be a contender once that networking is ready next year. And AMD can already out-perform the Nvidia H200, once you clear the ROCm development hurdle.

The Future of AI Hardware

It could take a year or more for AMD to be able to scale efficiently, and by then Nvidia will have moved on to the Kyber-based NVL576 with the new NVLink7, Vera CPU and upgraded Rubin GPU.

Conclusion

If you start late; you stay behind. The AI hardware market is rapidly evolving, and companies need to stay ahead of the curve to remain competitive.

FAQs

• What is MLPerf?
  MLPerf is a benchmarking suite for machine learning workloads, used to evaluate the performance of AI hardware.
• What is the difference between training and inference?
  Training refers to the process of training a machine learning model, while inference refers to the process of using a trained model to make predictions.
• What is NVLink?
  NVLink is a high-speed interconnect developed by Nvidia, used to connect GPUs and other devices in a system.
• What is UA-Link?
  UA-Link is a high-speed interconnect developed by AMD, used to connect GPUs and other devices in a system.
• What is ROCm?
  ROCm is an open-source software platform developed by AMD, used to manage and optimize machine learning workloads on AMD hardware.

Software and platforms for Diversity, Equity, Inclusion, and Accessibility (DEIA) are revolutionizing the way we approach social and economic disparities. By harnessing the power of technology, we can create a more just and equitable society for all. In this article, we’ll explore the critical role of software in shaping a more inclusive future.

The Current State of Inequality

The world is facing numerous challenges, from racial and gender disparities to unequal access to education and economic opportunities. These inequalities have far-reaching consequences, including social unrest, economic stagnation, and a decline in overall well-being. It’s essential to address these issues and create a more equitable society.

The Impact of Inequality on Society

Inequality affects not only individuals but also entire communities and societies. It can lead to social and economic instability, decreased economic growth, and a decline in mental and physical health. Furthermore, inequality can perpetuate cycles of poverty, making it challenging for marginalized groups to break free from systemic barriers.

The Role of Technology in Perpetuating Inequality

Technology can both perpetuate and alleviate inequality. On one hand, it can exacerbate existing disparities by providing unequal access to resources, information, and opportunities. On the other hand, technology can be a powerful tool for promoting inclusivity and equality. By developing and implementing inclusive software and platforms, we can create a more level playing field and provide opportunities for marginalized groups to thrive.

The Power of Inclusive Tech

Inclusive tech refers to software and platforms designed to promote diversity, equity, inclusion, and accessibility. These technologies can help address social and economic disparities by providing equal access to resources, information, and opportunities. Inclusive tech can take many forms, including accessible websites, mobile apps, and online platforms.

Examples of Inclusive Tech

There are numerous examples of inclusive tech, including:
– Accessible websites and mobile apps that provide equal access to information and resources for people with disabilities.
– Online platforms that promote diversity and inclusion in the workplace, such as diversity and inclusion training programs.
– Software that helps address systemic barriers, such as bias detection tools and diversity metrics.

The Benefits of Inclusive Tech

Inclusive tech offers numerous benefits, including:
– Increased diversity and inclusion in the workplace and society.
– Improved accessibility and equal access to resources and opportunities.
– Enhanced social and economic mobility for marginalized groups.
– A more equitable and just society.

Challenges and Opportunities

While inclusive tech has the potential to create a more equitable society, there are challenges and opportunities that must be addressed. These include:
– Ensuring equal access to technology and digital literacy.
– Addressing bias and discrimination in AI and machine learning algorithms.
– Developing and implementing inclusive tech that meets the needs of diverse users.

Addressing Bias and Discrimination

Bias and discrimination in AI and machine learning algorithms can perpetuate existing disparities and create new ones. It’s essential to develop and implement algorithms that are transparent, fair, and unbiased. This can be achieved by:
– Using diverse and representative data sets.
– Implementing bias detection and mitigation tools.
– Developing algorithms that prioritize fairness and equity.

Developing Inclusive Tech

Developing inclusive tech requires a deep understanding of the needs and experiences of diverse users. This can be achieved by:
– Conducting user research and testing.
– Involving diverse stakeholders in the development process.
– Prioritizing accessibility and usability.

Case Studies and Success Stories

There are numerous case studies and success stories that demonstrate the impact of inclusive tech. These include:
– Companies that have implemented diversity and inclusion training programs, resulting in increased diversity and inclusion in the workplace.
– Organizations that have developed accessible websites and mobile apps, providing equal access to information and resources for people with disabilities.
– Software that has helped address systemic barriers, such as bias detection tools and diversity metrics.

Lessons Learned

These case studies and success stories offer valuable lessons learned, including:
– The importance of prioritizing diversity, equity, inclusion, and accessibility in tech development.
– The need for ongoing testing and evaluation to ensure that inclusive tech is effective and meets the needs of diverse users.
– The potential for inclusive tech to create a more equitable and just society.

Conclusion

In conclusion, software and platforms for DEIA have the potential to create a more equitable and just society. By harnessing the power of technology, we can address social and economic disparities and promote diversity, equity, inclusion, and accessibility. It’s essential to prioritize inclusive tech and develop software and platforms that meet the needs of diverse users. By doing so, we can create a brighter future for all.

Frequently Asked Questions

What is inclusive tech?

Inclusive tech refers to software and platforms designed to promote diversity, equity, inclusion, and accessibility.

How can inclusive tech address social and economic disparities?

Inclusive tech can address social and economic disparities by providing equal access to resources, information, and opportunities.

What are some examples of inclusive tech?

Examples of inclusive tech include accessible websites and mobile apps, online platforms that promote diversity and inclusion in the workplace, and software that helps address systemic barriers.

How can we ensure that inclusive tech is effective and meets the needs of diverse users?

We can ensure that inclusive tech is effective and meets the needs of diverse users by conducting user research and testing, involving diverse stakeholders in the development process, and prioritizing accessibility and usability.

What are some challenges and opportunities in developing and implementing inclusive tech?

Challenges and opportunities in developing and implementing inclusive tech include ensuring equal access to technology and digital literacy, addressing bias and discrimination in AI and machine learning algorithms, and developing and implementing inclusive tech that meets the needs of diverse users.

Introduction to AI Job Loss

Anthropic CEO Dario Amodei issued a warning last month that landed like a thunderclap in Silicon Valley and beyond. In what sounded almost like an apocalyptic future for workers around the globe, the 42-year billionaire predicted in a CNN interview with Anderson Cooper that within five years, AI could automate away up to 50% of all entry-level white-collar jobs.

It was a jarring prediction, even for an industry accustomed to provocative soundbites, especially coming from the head of the AI company behind Claude. The quote quickly ricocheted across news outlets, igniting headlines and debates about the economic future of billions. CNN, notably, cast the comments in a more skeptical light, asking whether dire forecasts about AI are becoming self-fulfilling. Others, like Axios, highlighted the fear among young professionals who are just beginning to understand how automation might shadow their careers.

The Reality of AI Automation

Experts across telecom, software and enterprise architecture suggest a more nuanced reality. Yes, AI is changing work — faster than ever before. But this isn’t just a story of job loss. It’s also about reinvention, overcorrection and the uniquely human skills machines still struggle to replicate.

An Unprecedented Pace Of Change

“Any industrial or technology revolution results in job loss. This has happened many times over,” said Andy Thurai, Field CTO at Cisco, in an interview. “What’s different this time is the speed. The AI hype cycle is moving much faster than anything we’ve seen before.”

Dima Gutzeit, founder and CEO of LeapXpert, echoed this sentiment. “We’re entering a high-speed workforce transformation,” he told me. “What’s different this time? The pace. Automation used to take decades — now it’s happening in quarters.”

Counting The Cost

Klarna made headlines in 2024 when it replaced 700 customer support agents with an AI chatbot. But it quietly brought back some of those roles in early 2025, realizing customers preferred human support to AI. Why? Because the bots weren’t flawless, as industry experts continue to warn.

Many companies are trimming senior teams and hoping AI-enhanced mid-level hires can close the gap. But just like in Klarna’s case, it’s not always working. “The results have been mixed so far,” said Thurai. “The pendulum always swings wide. Companies get seduced by cost savings and forget about institutional memory and strategic insight.”

The New Normal: A Hybrid Approach

Nowhere is this more evident than in telecom. Arnd Baranowski, founder and CEO of Oculeus, explained that while AI has become essential to fraud detection, it still needs human judgment.

“AI allows telecom providers to analyze massive volumes of traffic well beyond human capacity," Baranowski said. "But when fraudsters adopt unpredictable new methods, only humans can anticipate the shift. That requires imagination — and that’s something AI lacks.”

Between Alarm And Opportunity

Thurai believes many of the more dramatic claims from AI vendors serve a strategic purpose. “Obviously, the AI providers — Anthropic, OpenAI, consultants — have to say extreme things to gain attention and instill FOMO,” he said. “But there are people like IBM’s CEO with a more realistic picture of the future.”

Yes, AI will cause job losses. But it will also create roles — including data scientists, prompt engineers, AI governance experts — that didn’t exist five years ago.

So Will AI Take Your Job?

Maybe not. But the person who knows how to use it might. The big message from the experts is for global workers to move beyond the realm of FOMO into really understanding how to leverage AI tools for improved efficiency.

As Avanes put it: “AI isn’t here to optimize systems. It’s here to free people to focus on what matters. The question is whether we’ll let it.”

For Gutzeit, this an urgent call to reskill the global workforce. “The traditional career ladder is being cut off at the bottom. If we don’t reskill aggressively, we risk locking out an entire generation from meaningful career starts,” he said.

Conclusion

The impact of AI on the job market is a complex issue, and while there will be job losses, there will also be new opportunities and roles created. It is essential for workers to develop skills that complement AI and for organizations to adopt a hybrid approach that combines the benefits of AI with human judgment and expertise.

FAQs

Q: Will AI replace all jobs?
A: No, while AI will automate some jobs, it will also create new roles and opportunities.
Q: How can workers prepare for the changing job market?
A: Workers should focus on developing skills that complement AI, such as critical thinking, creativity, and problem-solving.
Q: What is the biggest challenge for organizations adopting AI?
A: The biggest challenge is finding the right balance between automation and human judgment, and ensuring that AI is used to augment human capabilities, not replace them.
Q: Will AI lead to significant job losses?
A: Yes, AI will lead to job losses, but it will also create new opportunities and roles, and it is essential for workers and organizations to adapt to the changing job market.