TechGiants Decode

A Publication of Tech Analysis

[TechGiants Decode @tgdc_miula]

Any argument built on the premise that “the AI race is already over,” or that one company is permanently ahead while another is doomed, can basically be skipped.

It has very little analytical value.

The AI race is still in the top of the third inning. Any lead can still be reversed.

Just look at the past few months.

Google’s Gemini 3.0 was hailed as the “endgame winner” late last year. By February, Anthropic’s Claude 4.6 had outperformed it in real-world work environments.

Then Anthropic, with Claude Opus 4.6, was viewed as having a decisive lead in agentic workflows for roughly three months. Some people started acting as if the AI game was already settled and Anthropic had become the inevitable winner.

But by late April, OpenAI managed to overtake again with GPT-5.5.

This kind of leapfrogging has already happened several times over the past two years. My guess is that it will continue for the next few years as well.

So we really should not rebuild our entire worldview just because one AI lab currently has the strongest model or product.

I still believe Meta and SpaceXAI both have meaningful opportunities ahead.

Even Microsoft’s MAI and Amazon’s Nova could eventually become serious contenders in this race.

#AI #ArtificialIntelligence #OpenAI #Anthropic #Google #Meta #Microsoft #Amazon #SpaceXAI #GenerativeAI #AgenticAI #LLM #TechStrategy #AIStrategy

[TechGiants Decode @tgdc_miula]

Google’s Services operating margins just proved it: The biggest worry of the past two years — that AI would blow up search costs — is now officially history.

Four factors let Google break through the business model wall:

• The original cost base was already tiny
• Inference costs have collapsed
• AI is pulling away the low-monetization queries
• AI is dramatically lifting monetization efficiency on everything else

The feared margin collapse never happened. Instead, search is getting stronger.

#Google #AI #Search

[TechGiants Decode @tgdc_miula]

Very bullish on what Mark Zuckerberg said about AI during the call.
#META

[TechGiants Decode @tgdc_miula]

Intel needs best process tech to make best product. If everyone use the same TSMC node, Intel maybe 1% better or 1 % worse than AMD, that means bloody market fight and not a great future.

[TechGiants Decode @tgdc_miula]

@herbgreenberg @ParagonIntelNYC It will likely take at least two more years for these advancements to translate into improved financial results. It seems that the board lacks the patience to wait for this transition to be fully realized.

[TechGiants Decode @tgdc_miula]

Under Pat Gelsinger’s leadership, Intel has managed to push forward its process technology development. Both the Intel 4 and Intel 3 node are now online and demonstrate considerable competitiveness, only slightly lagging behind TSMC's N3 node. Additionally, the 18A node is expected to come online smoothly next year, with the capability to compete with TSMC’s N2.

[TechGiants Decode @tgdc_miula]

Under Pat Gelsinger’s leadership, Intel has managed to push forward its process technology development. Both the Intel 4 and Intel 3 node are now online and demonstrate considerable competitiveness, only slightly lagging behind TSMC's N3 node. Additionally, the 18A node is expected to come online smoothly next year, with the capability to compete with TSMC’s N2.

However, it will likely take at least two more years for these advancements to translate into improved financial results. It seems that the board lacks the patience to wait for this transition to be fully realized.

The question is, does the board have a better solution to offer? Let’s wait and see. #PatGelsinger #Intel

[TechGiants Decode @tgdc_miula]

I’m actually quite curious why Intel decided not to pursue the integrated memory packaging approach of Lunar Lake. Of course, buying memory to integrate it would indeed lower the gross margin, but that’s just a superficial figure; actual profitability would not decrease.

For example, if the original CPU is sold for 10,000 with a cost of 5,000, the gross margin is 50%, and the gross profit is 5,000. Now, if memory is integrated, assuming the cost of 32G memory wafer is 2,000, and Intel makes no additional profit on it, the product price would increase to 12,000, and the cost would rise to 7,000. The gross margin would drop to 41.67%, but the actual profit would remain at 5,000, with no reduction.

Clearly, integrating memory would help with power consumption and latency. Apart from making the gross margin percentage look worse, there’s no real operational downside. I’m curious why Intel considers this experiment a failure and doesn’t plan to continue with it. Is there any expert willing to share their thoughts? #INTEL

[TechGiants Decode @tgdc_miula]

The Wall Street Journal reported that Qualcomm recently made an acquisition offer to Intel, but it is currently unclear whether Intel is willing to agree to the merger. Without discussing Intel’s willingness, I believe the biggest issue with this deal is that there doesn’t seem to be much potential for synergy after the integration. Qualcomm acquiring Intel may not result in a "1+1 greater than 2" scenario.

In terms of chip design, Qualcomm can’t really help Intel, and when it comes to chip manufacturing, Qualcomm is even less able to assist. I don’t see any compelling reason for Intel to agree to this acquisition. On the contrary, Qualcomm’s rationale is more understandable—they’re trying to take advantage of a bargain to enter the PC/server market that they’ve long struggled to break into.

In my opinion, there’s only one situation in which Intel would agree to be acquired: if the progress of its 18A process is going very poorly. So if Intel does express willingness to be acquired, it would actually be a bad sign.

$INTC

[TechGiants Decode @tgdc_miula]

Under Andy Jassy’s leadership, Amazon has demonstrated remarkable operational capabilities and a clear strategic direction. The company has been bold in making operational changes and investing in the future. All of this currently looks very positive, and it’s likely to lead Amazon to a better future—a potential that has clearly been underestimated by the public.
$AMZN #AMZN

[TechGiants Decode @tgdc_miula]

The overall cooling of the advertising market doesn't seem too bad. When Google released its earnings report, YouTube's growth rate slowed down too quickly. At that time, I thought it might be possible that the overall advertising market had slightly cooled. However, with META's earnings report this time, I feel that this concern can be alleviated. The overall advertising market still appears to be quite strong.
#META

[Sean Kelly @seanpk]

Elon Musk fired 80% of Twitter (6500 people) and everyone thought that Twitter was doomed.

He was right. Everyone was wrong.

It’s the management masterclass of the decade and every entrepreneur must understand why it worked 🧵:

[TechGiants Decode @tgdc_miula]

Quick Review of Apple's WWDC: Not Many Surprises, But Enough for Apple

After quickly reviewing the WWDC reports, I concluded that while Apple's AI-related announcements didn't offer many surprises, they might be sufficient purely from a product sales perspective. I believe the current generation iPhone 16 will likely trigger an upgrade wave.

Regarding the AI announcements, Apple essentially introduced a system where a small proprietary model runs on the phone's edge, while more complex tasks are handled in the cloud by a larger model, which currently seems to be a combination of Apple's proprietary large model and OpenAI's GPT-4o. Additionally, Apple can integrate this with Siri to link various app functions on the phone. Strictly speaking, none of these concepts are new. Microsoft had already integrated AI models and API calls at the OS level with Copilot+ for PCs, so Apple isn't bringing anything groundbreaking here. The real test will be whether Apple's small model can outperform Microsoft's Phi-3 when it officially launches.

However, this level of AI advancement might just be enough for Apple. After all, Apple's main business is selling hardware. Previously launched iPhones, except for the iPhone 15 Pro series, cannot upgrade to Apple Intelligence—even the latest iPhone 15 bought just two months ago. This limitation could significantly boost sales by attracting iPhone 12 and iPhone 13 users to upgrade. Having AI capabilities will undoubtedly enhance user experience, providing sufficient motivation for consumers to upgrade.

Will external doubts about Apple's lag in the AI wave disappear? Probably not. Apple's advantage of having a vast number of end devices is always there, similar to the unparalleled distribution network of convenience stores—always the strongest channel for AI applications. Apple might continue to excel as a later adopter providing a better-integrated experience, which they have always been good at. But this doesn't change the fact that Apple is lagging in core AI technologies.

Of course, we can't predict how the leading or lagging in large language model technology will impact the business landscape in five to ten years—this is a question for astrologers. However, for tech giants, being a leader is always preferable to being a follower. I don't believe Apple's current position is where they want to be. Technological leadership is a competitive advantage, while technological lag is a business risk. This fundamental truth won't change.

In the short term, I believe Apple's next-generation phones should sell very well, ending the lackluster performance of the past two generations. After all, comparing the current Siri to the future Siri, you can see the reason.

#WWDC
#APPLE

[TechGiants Decode @tgdc_miula]

If Musk had initially invested one billion in xAI, and it grew tenfold within a year, that would truly demonstrate his capability as a wealth creation machine. Despite Tesla's challenges over the past two years, never underestimate Musk.

[TechGiants Decode @tgdc_miula]

Originally, I predicted a two-year low period for Tesla, but now it seems it could be shortened to one and a half years. Tesla appears to have made some realistic adjustments to the $25K Model, allowing it to share production lines with existing models and launch sooner. This decision, though contrary to Musk's usual approach, is the right one. Additionally, the progress on the robots aligns with my expectations, making a 2025 launch quite feasible. $TSLA #Tesla #Musk

[TechGiants Decode @tgdc_miula]

@elonmusk and the prompt worked last year won't work this year

[TechGiants Decode @tgdc_miula]

@lexfridman The problem with Atlas is not body but brain.

[TechGiants Decode @tgdc_miula]

A group of Google employees occupied the office of Google Cloud CEO Thomas Kurian to protest against Google's cloud service contract with the Israeli government. The protesters were arrested after refusing to comply with police orders to leave. Reports indicate that these employees have been terminated by Google. This incident highlights a period of intense internal conflict at Google, as top management's patience wears thin with highly progressive activist employees.
#GoogleConflict #CorporateActivism